WO2015135344A1 - 检测音频信号的方法和装置 - Google Patents

检测音频信号的方法和装置 Download PDF

Info

Publication number
WO2015135344A1
WO2015135344A1 PCT/CN2014/092694 CN2014092694W WO2015135344A1 WO 2015135344 A1 WO2015135344 A1 WO 2015135344A1 CN 2014092694 W CN2014092694 W CN 2014092694W WO 2015135344 A1 WO2015135344 A1 WO 2015135344A1
Authority
WO
WIPO (PCT)
Prior art keywords
audio signal
determined
ssnr
snr
signal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2014/092694
Other languages
English (en)
French (fr)
Inventor
王喆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Original Assignee
Huawei Technologies Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to CA2940487A priority Critical patent/CA2940487C/en
Priority to JP2016556770A priority patent/JP6493889B2/ja
Priority to EP14885786.5A priority patent/EP3118852B1/en
Priority to RU2016139717A priority patent/RU2666337C2/ru
Priority to KR1020187021506A priority patent/KR102005009B1/ko
Priority to EP19197660.4A priority patent/EP3660845B1/en
Priority to AU2014386442A priority patent/AU2014386442B9/en
Priority to SG11201607052SA priority patent/SG11201607052SA/en
Priority to MYPI2016703030A priority patent/MY193521A/en
Application filed by Huawei Technologies Co Ltd filed Critical Huawei Technologies Co Ltd
Priority to MX2016011750A priority patent/MX355828B/es
Priority to KR1020167025280A priority patent/KR101884220B1/ko
Priority to ES14885786T priority patent/ES2787894T3/es
Publication of WO2015135344A1 publication Critical patent/WO2015135344A1/zh
Priority to US15/262,263 priority patent/US10304478B2/en
Anticipated expiration legal-status Critical
Priority to US16/391,893 priority patent/US10818313B2/en
Priority to US16/901,846 priority patent/US11417353B2/en
Ceased legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/78Detection of presence or absence of voice signals
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00Speech recognition
    • G10L15/02Feature extraction for speech recognition; Selection of recognition unit
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00Speech recognition
    • G10L15/08Speech classification or search
    • G10L15/10Speech classification or search using distance or distortion measures between unknown speech and reference templates
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00Speech recognition
    • G10L15/20Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/93Discriminating between voiced and unvoiced parts of speech signals
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/78Detection of presence or absence of voice signals
    • G10L2025/783Detection of presence or absence of voice signals based on threshold decision
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/18Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band

Definitions

  • Embodiments of the present invention relate to the field of signal processing techniques and, more particularly, to methods and apparatus for detecting audio signals.
  • VAD Voice Activity Detection
  • SAD Sound Activity Detection
  • Typical activity signals include voice, music, and the like.
  • the principle of VAD is to extract one or more feature parameters from the input audio signal, determine one or more feature values according to the one or more feature parameters, and then combine the one or more feature values with one or more thresholds. Values are compared.
  • a segmentation signal to noise ratio (SSNR)-based active signal detection method in the prior art divides an input audio signal into frequency bands into a plurality of sub-band signals, and calculates the audio signal in each sub-band.
  • the energy is obtained by comparing the energy of the audio signal in each sub-band with the energy of an estimated background noise signal in each sub-band to obtain a signal-to-noise ratio of the audio signal on each sub-band (Signal-to- Noise Ratio, SNR).
  • determining SSNR according to the subband SNR on each subband comparing the SSNR with a preset VAD decision threshold, if the SSNR exceeds the VAD decision threshold, the audio signal is an active signal; if the SSNR does not exceed the VAD decision The threshold is the inactive signal.
  • a typical way to calculate the SSNR is to add all the sub-band SNRs of the audio signal, and the result is the SSNR.
  • the SSNR can be determined using Equation 1.1:
  • k denotes the kth subband
  • snr(k) denotes the subband SNR of the kth subband
  • N denotes the number of subbands in which the audio signal is divided into subbands in total.
  • the missed detection of the active speech may be caused.
  • Embodiments of the present invention provide a method and apparatus for detecting an audio signal capable of accurately distinguishing between active speech and inactive speech.
  • an embodiment of the present invention provides a method for detecting an audio signal, the method comprising: determining an input audio signal as an audio signal to be determined; determining an enhanced segmentation signal to noise ratio (SSNR) of the audio signal, where the enhanced SSNR is greater than Baseline SSNR; comparing the enhanced SSNR with a voice activity detection VAD decision threshold to determine if the audio signal is an active signal.
  • SSNR segmentation signal to noise ratio
  • the determined input audio signal is an audio signal to be determined, including: determining, according to a subband SNR of the audio signal, the audio signal is The audio signal is to be judged.
  • the determined input audio signal is an audio signal to be determined, including: the sub-band SNR is greater than the audio signal
  • the audio signal is determined to be the audio signal to be determined.
  • the determining the input audio signal is an audio signal to be determined, including: the sub-band SNR is greater in the audio signal Determining that the audio signal is to be determined if the number of high frequency terminal strips of the first preset threshold is greater than the second number and the number of low frequency terminal strips of the audio signal in which the subband SNR is less than the second preset threshold is greater than the third number audio signal.
  • the determining the input audio signal is an audio signal to be determined, including: a neutron band SNR in the audio signal
  • the audio signal is determined to be the audio signal to be determined.
  • the determined input audio signal is an audio signal to be determined, including: determining that the audio signal is an unvoiced signal, determining that the audio signal is The audio signal is to be judged.
  • the determining the enhanced segmentation signal to noise ratio of the audio signal includes: determining a weight of a subband SNR of each subband in the audio signal, wherein the subband SNR of the subband SNR greater than the first preset threshold has a weight of a subband SNR greater than that of the other subbands Weighting; determining the enhanced SSNR based on the weight of the subband SNR of each subband in the audio signal and the subband SNR of each subband.
  • determining the enhanced segmentation signal to noise ratio SSNR of the audio signal comprises: determining a reference SSNR of the audio signal; and determining an enhanced SSNR according to a reference SSNR of the audio signal.
  • the comparing the enhanced SSNR with the voice activity detection VAD determination threshold further includes: The VAD decision threshold is reduced by using a preset algorithm, and the reduced VAD decision threshold is obtained.
  • the enhanced SSNR is compared with the voice activity detection VAD decision threshold, and determining whether the audio signal is an active signal specifically includes: the enhanced SSNR and the The reduced VAD decision threshold is compared to determine if the audio signal is an active signal.
  • an embodiment of the present invention provides a method for detecting an audio signal, the method comprising: determining an input audio signal as an audio signal to be determined; determining a weight of a sub-band signal to noise ratio SNR of each subband in the audio signal, The weight of the subband SNR of the high frequency terminal strip with the subband SNR greater than the first preset threshold is greater than the weight of the subband SNR of the other subbands; the weight and each of the subband SNR according to each subband in the audio signal
  • the subband SNR of the subband determines an enhanced segmentation signal to noise ratio SSNR, wherein the enhanced SSNR is greater than a reference SSNR; the enhanced SSNR is compared to a voice activity detection VAD decision threshold to determine whether the audio signal is an active signal.
  • the determining input The audio signal is an audio signal to be determined, and includes: determining, according to a sub-band SNR of the audio signal, the audio signal as an audio signal to be determined.
  • the determined input audio signal is an audio signal to be determined, including: the sub-band SNR is greater than the audio signal
  • the audio signal is determined to be the audio signal to be determined.
  • the determined input audio signal is an audio signal to be determined, including: the sub-band SNR is greater than the audio signal Determining that the audio signal is to be determined if the number of high frequency terminal strips of the first preset threshold is greater than the second number and the number of low frequency terminal strips of the audio signal in which the subband SNR is less than the second preset threshold is greater than the third number audio signal.
  • an embodiment of the present invention provides a method for detecting an audio signal, the method comprising: determining an input audio signal as an audio signal to be determined; acquiring a reference segmental signal to noise ratio SSNR of the audio signal; using a preset algorithm to reduce The small reference voice activity detects the VAD decision threshold, obtains the reduced VAD decision threshold, compares the reference SSNR with the reduced VAD decision threshold, and determines whether the audio signal is an active signal.
  • the determined input audio signal is an audio signal to be determined, including: determining, according to a subband SNR of the audio signal, the audio signal is The audio signal is to be judged.
  • the determined input audio signal is an audio signal to be determined, including: the sub-band SNR is greater in the audio signal When the number of the high frequency terminal strips of the first preset threshold is greater than the first number, the audio signal is determined to be the audio signal to be determined.
  • the determined input audio signal is an audio signal to be determined, including: the sub-band SNR is greater in the audio signal Determining that the audio signal is to be determined if the number of high frequency terminal strips of the first preset threshold is greater than the second number and the number of low frequency terminal strips of the audio signal in which the subband SNR is less than the second preset threshold is greater than the third number audio signal.
  • the determined input audio signal is an audio signal to be determined, including: a neutron band SNR in the audio signal
  • the audio signal is determined to be an audio signal to be determined.
  • the determined input audio signal is an audio signal to be determined, including: determining that the audio signal is an unvoiced signal, determining that the audio signal is The audio signal is to be judged.
  • an embodiment of the present invention provides an apparatus, where the apparatus includes: a first determining unit, configured to determine an input audio signal as an audio signal to be determined; and a second determining unit, configured to determine an enhanced segment of the audio signal a signal-to-noise ratio SSNR, wherein the enhanced SSNR is greater than a reference SSNR; and a third determining unit configured to compare the enhanced SSNR with a voice activity detection VAD decision threshold to determine whether the audio signal is an active signal.
  • the first determining unit is specifically configured to determine, according to a subband SNR of the audio signal, the audio signal as an audio signal to be determined.
  • the first determining unit is configured to: in the audio signal, the sub-band signal-to-noise ratio SNR is greater than the first When the number of high frequency terminal strips of the preset threshold is greater than the first number, the audio signal is determined to be an audio signal to be determined.
  • the first determining unit is configured to: in the audio signal, the subband SNR is greater than the first preset threshold In the case where the number of high frequency terminal strips is greater than the second number and the number of low frequency terminal strips in which the subband SNR is less than the second preset threshold in the audio signal is greater than the third number, the audio signal is determined to be the audio signal to be determined.
  • the first determining unit is configured to use, in the audio signal, that the value of the sub-band SNR is greater than the third pre- In the case where the number of sub-bands of the threshold is greater than the fourth number, the audio signal is determined to be the audio signal to be determined.
  • the first determining unit is configured to determine, in the case that the audio signal is an unvoiced signal, the audio signal is an audio signal to be determined.
  • the second determining unit is specifically configured to determine the The weight of the subband SNR of each subband in the audio signal, wherein the subband SNR is greater than The weight of the sub-band SNR of the high-frequency terminal strip of a preset threshold is greater than the weight of the sub-band SNR of the other sub-bands, according to the weight of the sub-band SNR of each sub-band in the audio signal and the sub-band SNR of each sub-band, Determine the enhanced SSNR.
  • the second determining unit is specifically configured to determine a reference SSNR of the audio signal, and determine the enhanced SSNR according to a reference SSNR of the audio signal.
  • the device further includes a fourth determining unit, where the fourth determining unit is configured to use The preset algorithm reduces the VAD decision threshold, and obtains the reduced VAD decision threshold.
  • the third determining unit is specifically configured to compare the enhanced SSNR with the reduced VAD decision threshold to determine whether the audio signal is Activity signal.
  • an embodiment of the present invention provides an apparatus, where the apparatus includes: a first determining unit, configured to determine an input audio signal as an audio signal to be determined; and a second determining unit, configured to determine each subband in the audio signal
  • the weight of the sub-band signal-to-noise ratio SNR wherein the sub-band SNR is greater than the weight of the sub-band SNR of the high-frequency terminal strip of the first preset threshold, and the weight of the sub-band SNR of the other sub-bands is greater, according to each of the audio signals
  • the weight of the subband SNR of the subband and the subband SNR of each subband determine an enhanced segmentation signal to noise ratio SSNR, wherein the enhanced SSNR is greater than a reference SSNR; and a third determining unit for detecting the enhanced SSNR and the voice activity detection VAD
  • the decision threshold is compared to determine whether the audio signal is an active signal.
  • the first determining unit is configured to determine, according to a subband SNR of the audio signal, the audio signal as an audio signal to be determined.
  • the first determining unit is configured to: in the audio signal, the sub-band signal-to-noise ratio SNR is greater than the first When the number of high frequency terminal strips of the preset threshold is greater than the first number, the audio signal is determined to be an audio signal to be determined.
  • the first determining unit is configured to: in the audio signal, the subband SNR is greater than the first preset threshold In the case where the number of high frequency terminal strips is greater than the second number and the number of low frequency terminal strips in which the subband SNR is less than the second preset threshold in the audio signal is greater than the third number, the audio signal is determined to be the audio signal to be determined.
  • an embodiment of the present invention provides an apparatus, where the apparatus includes: a first determining unit, configured to determine an input audio signal as an audio signal to be determined; and a second determining unit, configured to acquire a reference segment of the audio signal a signal-to-noise ratio SSNR; a third determining unit, configured to reduce a reference voice activity detection VAD decision threshold by using a preset algorithm to obtain a reduced VAD decision threshold; and a fourth determining unit, configured to reduce the reference SSNR and the reduction The subsequent VAD decision threshold is compared to determine if the audio signal is an active signal.
  • the first determining unit is configured to determine, according to the sub-band signal-to-noise ratio SNR of the audio signal, the audio signal as the to-be-determined audio signal.
  • the first determining unit is configured to: in the audio signal, the subband SNR is greater than the first preset threshold In the case where the number of high frequency terminal strips is greater than the first number, the audio signal is determined to be the audio signal to be determined.
  • the first determining unit is configured to: in the audio signal, the subband SNR is greater than the first preset threshold In the case where the number of high frequency terminal strips is greater than the second number and the number of low frequency terminal strips in which the subband SNR is less than the second preset threshold in the audio signal is greater than the third number, the audio signal is determined to be the audio signal to be determined.
  • the first determining unit is configured to: in the audio signal, the value of the neutron band SNR is greater than the third In a case where the number of sub-bands of the preset threshold is greater than the fourth number, the audio signal is determined to be an audio signal to be determined.
  • the first determining unit is configured to determine, in the case that the audio signal is an unvoiced signal, the audio signal as an audio signal to be determined.
  • the characteristics of the audio signal may be determined, and according to the characteristics of the audio signal, the enhanced SSNR is determined in a corresponding manner, and the enhanced SSNR is compared with the VAD decision threshold, so that the active signal is leaked.
  • FIG. 1 is a schematic flowchart of a method for detecting an audio signal according to an embodiment of the present invention.
  • FIG. 2 is a schematic flowchart of a method for detecting an audio signal according to an embodiment of the present invention.
  • FIG. 3 is a schematic flowchart of a method for detecting an audio signal according to an embodiment of the present invention.
  • FIG. 4 is a schematic flowchart of a method for detecting an audio signal according to an embodiment of the present invention.
  • FIG. 5 is a structural block diagram of an apparatus according to an embodiment of the present invention.
  • FIG. 6 is a structural block diagram of another apparatus according to an embodiment of the present invention.
  • FIG. 7 is a structural block diagram of an apparatus according to an embodiment of the present invention.
  • FIG. 8 is a structural block diagram of another apparatus according to an embodiment of the present invention.
  • FIG. 9 is a structural block diagram of another apparatus according to an embodiment of the present invention.
  • FIG. 10 is a structural block diagram of another apparatus according to an embodiment of the present invention.
  • FIG. 1 is a schematic flowchart of a method for detecting an audio signal according to an embodiment of the present invention.
  • the reference VAD decision threshold may be used when comparing the enhanced SSNR with the VAD decision threshold, or the reduced VAD decision threshold obtained after the reference VAD decision threshold may be reduced using a preset algorithm.
  • the reference VAD decision threshold may be a default VAD decision threshold.
  • the reference VAD decision threshold may be pre-stored or temporarily calculated. The calculation of the reference VAD decision threshold may be performed by using a prior art.
  • the preset algorithm may be to multiply the reference VAD decision threshold by a coefficient less than one, and other algorithms may be used.
  • the embodiment of the present invention does not limit the specific algorithm used. .
  • the SSNR of these audio signals may be lower than the preset VAD decision threshold.
  • these audio signals are active audio signals. This is due to the characteristics of these audio signals.
  • the sub-band SNR of the high frequency portion is significantly reduced.
  • the sub-band SNR of the high-frequency portion contributes less to the SSNR.
  • the SSNR calculated by the conventional SSNR calculation method may be lower than the VAD decision threshold, which causes the missed detection of the active signal.
  • the energy of the audio signal is relatively flat on the spectrum, but the overall energy of the audio signal is low.
  • the SSNR calculated using the conventional SSNR calculation method may also be lower than the VAD decision threshold. The method shown in FIG. 1 can effectively reduce the ratio of active signal leakage by appropriately increasing the SSNR such that the SSNR can be greater than the VAD decision threshold.
  • FIG. 2 is a schematic flowchart of a method for detecting an audio signal according to an embodiment of the present invention.
  • the spectrum of the input audio signal is divided into N subbands, where N is a positive integer greater than one.
  • the spectrum of the audio signal can be divided by psychoacoustic theory.
  • the width of the sub-band closer to the low frequency is narrower, and the width of the sub-band closer to the high frequency is wider.
  • the spectrum of the audio signal may be divided in other ways, for example, the spectrum of the audio signal is equally divided into N sub-bands.
  • a subband SNR is calculated for each subband of the input audio signal, wherein the subband SNR is the ratio of the energy of the subband to the energy of the background noise on the subband.
  • the subband energy of the background noise is generally estimated by the background noise estimator. estimated value. How to estimate the background noise energy corresponding to each sub-band by using the background noise estimator is well known in the art, and therefore, it is not necessary to go into details here.
  • the sub-band SNR may be a direct energy ratio or other representation of the direct energy ratio, such as a log sub-band SNR.
  • the sub-band SNR can also be a sub-band SNR or other deformation after linear or nonlinear processing on the direct sub-band SNR. The following formula is the direct energy ratio of the subband SNR:
  • snr(k) represents the subband SNR of the kth subband
  • E(k) and En(k) represent the energy of the kth subband and the energy of the background noise on the kth subband, respectively.
  • the sub-band energy used to calculate the sub-band SNR can be either the energy of the input audio signal on the sub-band or the energy of the input audio signal on the sub-band to remove the background noise in the sub-band. The energy after the energy.
  • the calculation of SNR is as long as it does not deviate from the meaning of SNR.
  • determining the input audio signal as the to-be-determined audio signal includes: determining, according to the sub-band SNR of the audio signal determined in step 201, the audio signal as the to-be-determined audio signal.
  • the determined input audio signal is an audio signal to be determined, including: in the audio signal.
  • the audio signal is determined to be an audio signal to be determined.
  • the determined input audio signal is an audio signal to be determined, including: the audio signal If the number of the high frequency terminal strips with the SNR greater than the first preset threshold is greater than the second number and the number of low frequency terminal strips in the audio signal where the subband SNR is less than the second preset threshold is greater than the third number,
  • the audio signal is an audio signal to be determined.
  • the high frequency end and the low frequency end of one frame of the audio signal are relatively speaking, that is, the portion having a relatively high frequency is a high frequency end, and the portion having a relatively low frequency is a low frequency end.
  • the determined input audio signal is an audio signal to be determined.
  • the number includes: determining that the audio signal is an audio signal to be determined if a value of the sub-band SNR in the audio signal is greater than a third number of sub-bands of the third preset threshold.
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the third preset threshold is also obtained based on statistics. Specifically, the third preset threshold is determined from the sub-band SNR of the large number of noise signals such that the sub-band SNR of most of the sub-bands of the noise signals is less than the value.
  • the first quantity, the second quantity, the third quantity, and the fourth quantity are also obtained based on statistics.
  • the first quantity in a large number of noise-containing voice unvoiced sample frames, the number of sub-bands whose SNR of the high-frequency terminal strip is larger than the first preset threshold is determined, and the first quantity is determined therefrom, so that the voice unvoiced samples are made.
  • the majority of the subband SNR in the frame is greater than the first preset threshold.
  • the number of high frequency terminal strips is greater than the first number.
  • the method of obtaining the second quantity is similar to the method of obtaining the first quantity.
  • the second number may be the same as the first quantity, and the second quantity may also be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is counted less than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice unvoiced samples The majority of the subband SNR in the frame is less than the second preset threshold.
  • the number of low frequency terminal strips is greater than the third number.
  • the statistical sub-band SNR is smaller than the number of sub-bands of the third preset threshold, and the fourth quantity is determined therefrom, so that most of the sub-band SNRs of the noise sample frames are smaller than the third
  • the number of sub-bands of the preset threshold is greater than the fourth number.
  • whether the input audio signal is an audio signal to be determined may be determined by determining whether the input audio signal is an unvoiced signal. In this case, it is not necessary to determine the sub-band SNR of the audio signal when determining whether the audio signal is an audio signal to be judged. In other words, step 201 does not need to be performed in determining whether the audio signal is an audio signal to be determined. Specifically, the determining the input audio signal is the audio signal to be determined, and if the audio signal is determined to be an unvoiced signal, determining the audio signal as the audio signal to be determined. In particular, those skilled in the art will appreciate that there are a variety of methods for detecting whether an audio signal is an unvoiced signal.
  • the audio signal by detecting the Zero-Crossing Rate (ZCR) of the audio signal. Determine if the audio signal is an unvoiced signal. Specifically, in the case where the ZCR of the audio signal is greater than the ZCR threshold, the audio signal is determined to be an unvoiced signal, wherein the ZCR threshold is determined by a large number of experiments.
  • ZCR Zero-Crossing Rate
  • the reference SSNR can be the SSNR calculated using Equation 1.1. As can be seen from Equation 1.1, when calculating the reference SSNR, the subband SNR of any subband is not weighted, that is, the weight of the subband SNR of each subband is the same when calculating the reference SSNR.
  • the SSNR is enhanced, including: determining a weight of a subband SNR of each subband in the audio signal, where the weight of the high frequency terminal band of the subband SNR is greater than a weight of a subband SNR of other subbands according to a first preset threshold, according to The weight of the subband SNR of each subband in the audio signal and the subband SNR of each subband determine the enhanced SSNR.
  • the audio signal is divided into 20 sub-bands according to psychoacoustic theory, that is, sub-band 0 to sub-band 19.
  • the sub-band 18 and the sub-band 19 are both larger than the first predetermined value T1
  • four sub-bands that is, the sub-band 20 to the sub-band 23 may be added.
  • the sub-band 18 having a signal-to-noise ratio greater than T1 may be divided into a sub-band 18a, a sub-band 18b, and a sub-band 18c, and the sub-band 19 is divided into a sub-band 19a, a sub-band 19b, and a sub-band 19c.
  • the sub-band 18 can be regarded as a mother-sub-belt of the sub-band 18a, the sub-band 18b, and the sub-band 18c
  • the sub-band 19 can be regarded as a mother-child belt of the sub-band 19a, the sub-band 19b, and the sub-band 19c.
  • the values of the signal-to-noise ratio of the sub-band 18a, the sub-band 18b, and the sub-band 18c are the same as the values of the signal-to-noise ratio of the mother and sub-bands, and the values of the signal-to-noise ratio of the sub-band 19a, the sub-band 19b, and the sub-band 19c are the same as those of the mother-child band.
  • the noise ratio has the same value.
  • the original sub-divided 20 sub-bands are re-divided into 24 sub-bands. Since the VAD is still designed according to 20 subbands when performing active signal detection, it is necessary to map 24 subbands back to 20 subbands to determine the enhanced SSNR.
  • the enhanced SSNR is determined by increasing the number of high frequency terminal strips in which the subband SNR is greater than the first preset threshold, the following formula may be used for calculation:
  • SSNR' represents the enhanced SSNR.
  • Snr(k) represents the subband SNR of the kth subband.
  • the calculated reference SSNR is Obviously, the value of the enhanced SSNR calculated using Equation 1.3 for the first type of audio signal is greater than the value of the reference SSNR calculated using Equation 1.1.
  • the enhanced SSNR may be determined by the following formula:
  • SSNR indicates the reinforcing SSNR
  • snr (k) represents the k-th subband of the subband SNR
  • a 1 and a 2 to increase the weight parameter
  • a 1 and a value of 2 is such that a 1 ⁇ snr (18) + a 2 ⁇ snr(19) is larger than snr(18)+snr(19).
  • the value of the enhanced SSNR calculated using Equation 1.4 is greater than the value of the reference SSNR calculated using Equation 1.1.
  • determining the enhanced SSNR of the audio signal includes determining a reference SSNR of the audio signal, and determining an enhanced SSNR according to a reference SSNR of the audio signal.
  • the enhanced SSNR can be determined using the following formula:
  • SSNR represents the reference SSNR of the audio signal
  • SSNR' represents the enhanced SSNR
  • x and y represent the enhancement parameters.
  • the value of x can be 1.05
  • the value of y can be 1.
  • the values of x and y may also be other suitable values such that the enhanced SSNR is properly greater than the reference SSNR.
  • the enhanced SSNR can be determined using the following formula:
  • SSNR represents the original SSNR of the audio signal
  • SSNR' represents the enhanced SSNR
  • f(x), h(y) represents the enhancement function.
  • f(x) and h(y) may be functions related to the long-term SNR (LSNR) of the audio signal, and the long-term signal-to-noise ratio of the audio signal is for a long period of time.
  • Average SNR or weighted SNR For example, when lsnr is greater than 20, f(lsnr) may be equal to 1.1 and y(lsnr) may be equal to 2.
  • f(lsnr) When lsnr is less than 20 and greater than 15, f(lsnr) may be equal to 1.05, and y(lsnr) may be equal to 1. When lsnr is less than 15, f(lsnr) may be equal to 1, and y(lsnr) may be equal to zero.
  • f(x) and h(y) may also be in other suitable forms such that the enhanced SSNR is properly greater than the reference SSNR.
  • the enhanced SSNR is compared with a VAD decision threshold, and if the enhanced SSNR is greater than the VAD decision threshold, the audio signal is determined to be an active signal. Otherwise it is determined that the audio signal is an inactive signal.
  • the method may further include: reducing the VAD decision threshold by using a preset algorithm to obtain a reduced VAD decision threshold.
  • comparing the enhanced SSNR with the VAD decision threshold specifically includes comparing the enhanced SSNR with the reduced VAD decision threshold to determine whether the audio signal is an active signal.
  • the reference VAD decision threshold may be a default VAD decision threshold, which may be pre-stored or temporarily calculated, wherein the calculation of the reference VAD decision threshold may be performed by a prior art technique.
  • the preset algorithm may be to multiply the reference VAD decision threshold by a coefficient less than one, and other algorithms may be used.
  • the embodiment of the present invention does not limit the specific algorithm used. .
  • the preset algorithm may appropriately reduce the VAD decision threshold such that the enhanced SSNR is greater than the reduced VAD decision threshold, so that the proportion of the active signal being missed may be reduced.
  • the characteristics of the audio signal are determined, and according to the characteristics of the audio signal, the enhanced SSNR is determined in a corresponding manner, and the enhanced SSNR is compared with the VAD decision threshold, so that the active signal is reduced in the proportion of missed detection. .
  • FIG. 3 is a schematic flowchart of a method for detecting an audio signal according to an embodiment of the present invention.
  • the reference SSNR can be the SSNR calculated using Equation 1.1. As can be seen from Equation 1.1, when calculating the reference SSNR, the subband SNR of any subband is not weighted, that is, the weight of the subband SNR of each subband is the same when calculating the reference SSNR.
  • the audio signal is divided into 20 sub-bands according to psychoacoustic theory, that is, sub-band 0 to sub-band 19. If the sub-band 18 and the sub-band 19 are both larger than the first preset value T1, four more can be added.
  • Sub-bands that is, sub-bands 20 to sub-bands 23.
  • the sub-band 18 having a signal-to-noise ratio greater than T1 may be divided into a sub-band 18a, a sub-band 18b, and a sub-band 18c, and the sub-band 19 is divided into a sub-band 19a, a sub-band 19b, and a sub-band 19c.
  • the sub-band 18 can be regarded as a mother-sub-belt of the sub-band 18a, the sub-band 18b, and the sub-band 18c
  • the sub-band 19 can be regarded as a mother-child belt of the sub-band 19a, the sub-band 19b, and the sub-band 19c.
  • the values of the signal-to-noise ratio of the sub-band 18a, the sub-band 18b, and the sub-band 18c are the same as the values of the signal-to-noise ratio of the mother and sub-bands, and the values of the signal-to-noise ratio of the sub-band 19a, the sub-band 19b, and the sub-band 19c are the same as those of the mother-child band.
  • the noise ratio has the same value.
  • the original sub-divided 20 sub-bands are re-divided into 24 sub-bands. Since the VAD is still designed according to 20 subbands when performing active signal detection, it is necessary to map 24 subbands back to 20 subbands to determine the enhanced SSNR.
  • the enhanced SSNR is determined by increasing the number of high frequency terminal strips in which the subband SNR is greater than the first preset threshold, the following formula may be used for calculation:
  • SSNR' represents the enhanced SSNR.
  • Snr(k) represents the subband SNR of the kth subband.
  • the calculated reference SSNR is Obviously, the value of the enhanced SSNR calculated using Equation 1.3 for the first type of audio signal is greater than the value of the reference SSNR calculated using Equation 1.1.
  • the enhanced SSNR may be determined by the following formula:
  • SSNR indicates the reinforcing SSNR
  • snr (k) represents the k-th subband of the subband SNR
  • a 1 and a 2 to increase the weight parameter
  • a 1 and a value of 2 is such that a 1 ⁇ snr (18) + a 2 ⁇ snr(19) is larger than snr(18)+snr(19).
  • the value of the enhanced SSNR calculated using Equation 1.4 is greater than the value of the reference SSNR calculated using Equation 1.1.
  • the enhanced SSNR is compared with a VAD decision threshold, and if the enhanced SSNR is greater than the VAD decision threshold, the audio signal is determined to be an active signal. Otherwise determine the audio letter The number is an inactive signal.
  • the method described in FIG. 3 can determine the characteristics of the audio signal, determine the enhanced SSNR in a corresponding manner according to the characteristics of the audio signal, and compare the enhanced SSNR with the VAD decision threshold, so that the active signal can be reduced by the missed detection ratio.
  • determining the input audio signal is the audio signal to be determined, and determining, according to the sub-band SNR of the audio signal, the audio signal is the audio signal to be determined.
  • determining the audio signal as an audio signal to be determined includes: in the audio signal neutron In the case where the number of high frequency terminal strips having an SNR greater than the first preset threshold is greater than the first number, the audio signal is determined to be an audio signal to be determined.
  • determining the audio signal as the audio signal to be determined includes: in the audio signal If the number of high frequency terminal strips whose subband SNR is greater than the first preset threshold is greater than the second number and the number of low frequency terminal strips in the audio signal where the subband SNR is less than the second preset threshold is greater than the third number, The audio signal is the audio signal to be judged.
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the first quantity, the second quantity, and the third quantity are also obtained based on statistics.
  • the first quantity as an example, in a large number of noise-containing voice unvoiced sample frames, the number of sub-bands whose SNR of the high-frequency terminal strip is larger than the first preset threshold is determined, and the first quantity is determined therefrom, so that the voice unvoiced samples are made.
  • the majority of the subband SNR in the frame is greater than the first preset threshold.
  • the number of high frequency terminal strips is greater than the first number.
  • the method of obtaining the second quantity is similar to the method of obtaining the first quantity.
  • the second number may be the same as the first quantity, and the second quantity may also be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is counted less than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice unvoiced samples The majority of the subband SNR in the frame is less than the second preset threshold.
  • the number of low frequency terminal strips is greater than the third number.
  • the embodiment of Figures 1 to 3 determines the input audio signal by using an enhanced SSNR. No is the activity signal.
  • the method shown in FIG. 4 determines whether the input audio signal is an active signal by reducing the VAD decision threshold.
  • FIG. 4 is a schematic flowchart of a method for detecting an audio signal according to an embodiment of the present invention.
  • determining the input audio signal as the to-be-determined audio signal includes: determining, according to the sub-band SNR of the audio signal determined in step 201, the audio signal as the to-be-determined audio signal.
  • the determined input audio signal is an audio signal to be determined, including: in the audio signal.
  • the audio signal is determined to be an audio signal to be determined.
  • the determined input audio signal is an audio signal to be determined, including: the audio signal If the number of the high frequency terminal strips with the SNR greater than the first preset threshold is greater than the second number and the number of low frequency terminal strips in the audio signal where the subband SNR is less than the second preset threshold is greater than the third number, The audio signal is an audio signal to be determined.
  • the determined input audio signal is an audio signal to be determined, including: the audio signal
  • the value of the sub-band SNR is greater than the third preset threshold, the number of sub-bands is greater than the fourth number, and the audio signal is determined to be the audio signal to be determined.
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the third preset threshold is also obtained based on statistics. Specifically, the third preset threshold is determined from the sub-band SNR of the large number of noise signals such that the sub-band SNR of most of the sub-bands of the noise signals is less than the value.
  • the first quantity, the second quantity, the third quantity, and the fourth quantity are also obtained based on statistics. Taking the first quantity as an example, in a large number of voice-voiced unvoiced sample frames containing noise, the sub-frequency band of the high-frequency terminal strip is counted. The number of subbands with an SNR greater than the first preset threshold is determined from the first number, such that the majority of the subband SNRs of the voice unvoiced sample frames are greater than the first preset threshold, and the number of high frequency terminal strips is greater than the first Quantity.
  • the method of obtaining the second quantity is similar to the method of obtaining the first quantity.
  • the second number may be the same as the first quantity, and the second quantity may also be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is counted less than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice unvoiced samples The majority of the subband SNR in the frame is less than the second preset threshold.
  • the number of low frequency terminal strips is greater than the third number.
  • the statistical sub-band SNR is smaller than the number of sub-bands of the third preset threshold, and the fourth quantity is determined therefrom, so that most of the sub-band SNRs of the noise sample frames are smaller than the third
  • the number of sub-bands of the preset threshold is greater than the fourth number.
  • whether the input audio signal is an audio signal to be determined may be determined by determining whether the input audio signal is an unvoiced signal. In this case, it is not necessary to determine the sub-band SNR of the audio signal when determining whether the audio signal is an audio signal to be judged. In other words, step 201 does not need to be performed in determining whether the audio signal is an audio signal to be determined. Specifically, the determining the input audio signal is the audio signal to be determined, and if the audio signal is determined to be an unvoiced signal, determining the audio signal as the audio signal to be determined. In particular, those skilled in the art will appreciate that there are a variety of methods for detecting whether an audio signal is an unvoiced signal.
  • whether the audio signal is an unvoiced signal can be determined by detecting a Zero-Crossing Rate (ZCR) of the audio signal.
  • ZCR Zero-Crossing Rate
  • the audio signal is determined to be an unvoiced signal, wherein the ZCR threshold is determined by a large number of experiments.
  • the reference SSNR may be the SSNR calculated using Equation 1.1.
  • the reference VAD decision threshold may be a default VAD decision threshold, which may be pre-stored or temporarily calculated, wherein the calculation of the reference VAD decision threshold may be performed by using a prior art technique.
  • the preset algorithm may be to multiply the reference VAD decision threshold by a coefficient less than one, and other algorithms may be used.
  • the embodiment of the present invention does not limit the specific algorithm used. .
  • the preset algorithm may appropriately reduce the VAD decision threshold such that the enhanced SSNR is greater than the reduced VAD decision gate. Limit, so that the proportion of the active signal being missed can be reduced.
  • the SSNR of these audio signals may be lower than the preset VAD decision threshold.
  • these audio signals are active audio signals. This is due to the characteristics of these audio signals.
  • the sub-band SNR of the high frequency portion is significantly reduced.
  • the sub-band SNR of the high-frequency portion contributes less to the SSNR.
  • the SSNR calculated by the conventional SSNR calculation method may be lower than the VAD decision threshold, which causes the missed detection of the active signal.
  • the energy of the audio signal is relatively flat on the spectrum, but the overall energy of the audio signal is low.
  • the SSNR calculated using the conventional SSNR calculation method may also be lower than the VAD decision threshold. The method shown in FIG. 4 reduces the VSNR decision threshold, so that the SSNR calculated by the conventional SSNR calculation method is greater than the VAD decision threshold, so that the ratio of the active signal leakage can be effectively reduced.
  • FIG. 5 is a structural block diagram of an apparatus according to an embodiment of the present invention.
  • the apparatus shown in Figure 5 is capable of performing the various steps of Figure 1 or Figure 2.
  • the apparatus 500 includes a first determining unit 501, a second determining unit 502, and a third determining unit 503.
  • the first determining unit 501 is configured to determine that the input audio signal is an audio signal to be determined.
  • the second determining unit 502 is configured to determine an enhanced segmentation signal to noise ratio SSNR of the audio signal, where the enhanced SSNR is greater than a reference SSNR.
  • the third determining unit 503 is configured to compare the enhanced SSNR with a voice activity detection VAD decision threshold to determine whether the audio signal is an active signal.
  • the apparatus 500 shown in FIG. 5 can determine the characteristics of the input audio signal, determine the enhanced SSNR in a corresponding manner according to the characteristics of the audio signal, and compare the enhanced SSNR with the VAD decision threshold, so that the active signal can be missed. The ratio is reduced.
  • the first determining unit 501 is specifically configured to determine, according to the subband SNR of the audio signal, the audio signal as an audio signal to be determined.
  • the first determining unit 501 determines, according to the subband SNR of the audio signal, that the audio signal is an audio signal to be determined, the first determining unit 501, specifically In the case where the number of high frequency terminal strips in which the subband SNR is greater than the first preset threshold in the audio signal is greater than the first number, the audio signal is determined to be the audio signal to be determined.
  • the first determining unit 501 determines that the audio signal is an audio signal to be determined according to the subband SNR of the audio signal
  • the first determining unit 501 is specifically configured to use the audio signal. If the number of the high frequency terminal strips whose SNR is greater than the first preset threshold is greater than the second number and the number of low frequency terminal strips in the audio signal where the subband SNR is less than the second preset threshold is greater than the third quantity, The audio signal is the audio signal to be judged.
  • the first determining unit 501 determines that the audio signal is an audio signal to be determined according to the subband SNR of the audio signal
  • the first determining unit 501 is specifically configured to use the audio signal.
  • the value of the sub-band SNR is greater than the third preset threshold
  • the number of sub-bands is greater than the fourth number, and the audio signal is determined to be the audio signal to be determined.
  • the first determining unit 501 is specifically configured to determine, in the case that the audio signal is an unvoiced signal, the audio signal as an audio signal to be determined.
  • the audio signal is an unvoiced signal
  • whether the audio signal is an unvoiced signal can be determined by detecting a Zero-Crossing Rate (ZCR) of the audio signal.
  • ZCR Zero-Crossing Rate
  • the audio signal is determined to be an unvoiced signal, wherein the ZCR threshold is determined by a large number of experiments.
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the third preset threshold is also obtained based on statistics. Specifically, the third preset threshold is determined from the sub-band SNR of the large number of noise signals such that the sub-band SNR of most of the sub-bands of the noise signals is less than the value.
  • the first quantity, the second quantity, the third quantity, and the fourth quantity are also obtained based on statistics.
  • the first quantity in a large number of voice samples containing noise, the number of subbands whose SNR of the high frequency terminal strip is larger than the first preset threshold is determined, and the first quantity is determined therefrom, so that the voice samples are absolutely large.
  • a majority of the high frequency terminal strips having a SNR greater than the first predetermined threshold are greater than the first number.
  • the method of determining the second quantity is similar to the method of determining the first quantity.
  • the second number may be the same as the first quantity or may be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is calculated to be greater than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice samples are absolutely large.
  • a majority of the low frequency terminal band SNR greater than the second predetermined threshold is greater than the third number.
  • the statistical sub-band SNR is greater than the number of sub-bands of the third preset threshold, and the fourth quantity is determined therefrom, so that the majority of the voice samples are greater than the third preset.
  • the number of subband SNRs of the threshold is greater than the fourth number.
  • the second determining unit 502 is specifically configured to determine a weight of the subband SNR of each subband in the audio signal, where the subband SNR is greater than the first preset threshold, and the weight of the high frequency terminal strip is greater than that of the other subbands.
  • the weight with SNR is determined according to the weight of the subband SNR of each subband in the audio signal and the SNR of each subband.
  • the second determining unit 502 is specifically configured to determine a reference SSNR of the audio signal, and determine an enhanced SSNR according to a reference SSNR of the audio signal.
  • the reference SSNR can be the SSNR calculated using Equation 1.1.
  • the sub-band SNRs of the respective sub-bands included in the SSNR have the same weight in the SSNR.
  • the second determining unit 502 is specifically configured to determine the enhanced SSNR by using the following formula:
  • SSNR represents the reference SSNR
  • SSNR' represents the enhanced SSNR
  • x and y represent enhancement parameters.
  • the value of x can be 1.05
  • the value of y can be 1.
  • the values of x and y may also be other suitable values such that the enhanced SSNR is properly greater than the reference SSNR.
  • the second determining unit 502 is specifically configured to determine the enhanced SSNR by using the following formula:
  • SSNR represents the reference SSNR
  • SSNR' represents the enhanced SSNR
  • f(x), h(y) represents an enhancement function.
  • f(x) and h(y) may be functions related to the long-term SNR (LSNR) of the audio signal, and the long-term signal-to-noise ratio of the audio signal is for a long period of time.
  • Average SNR or weighted SNR For example, when lsnr is greater than 20, f(lsnr) may be equal to 1.1 and y(lsnr) may be equal to 2. When lsnr is less than 20 and greater than 15, f(lsnr) can be equal to 1.05. y(lsnr) can be equal to 1.
  • f(lsnr) When lsnr is less than 15, f(lsnr) may be equal to 1, and y(lsnr) may be equal to zero.
  • f(x) and h(y) may also be in other suitable forms such that the enhanced SSNR is properly greater than the reference SSNR.
  • the third determining unit 503 is specifically configured to compare the enhanced SSNR with a voice activity detection VAD decision threshold, and determine, according to the comparison structure, whether the audio signal is an active signal. Specifically, if the enhanced SSNR is greater than the VAD decision threshold, it is determined that the audio signal is an active signal. If the enhanced SSNR is less than the VAD decision threshold, then the audio signal is determined to be an inactive signal.
  • the reduced VAD decision threshold obtained after the reference VAD decision threshold is reduced may also be used by using a preset algorithm, and the reduced VAD decision threshold is used to determine whether the audio signal is an active signal.
  • the apparatus 500 may further include a fourth determining unit 504.
  • the fourth determining unit 504 is configured to reduce the VAD decision threshold by using a preset algorithm to obtain a reduced VAD decision threshold.
  • the third determining unit 503 is specifically configured to compare the enhanced SSNR with the reduced VAD decision threshold to determine whether the audio signal is an active signal.
  • FIG. 6 is a structural block diagram of another apparatus according to an embodiment of the present invention.
  • the apparatus shown in Figure 6 is capable of performing the various steps of Figure 3.
  • the apparatus 600 includes a first determining unit 601, a second determining unit 602, and a third determining unit 603.
  • the first determining unit 601 is configured to determine that the input audio signal is an audio signal to be determined.
  • a second determining unit 602 configured to determine a weight of a sub-band signal-to-noise ratio SNR of each sub-band in the audio signal, where the sub-band SNR is greater than a first sub-band of the high-frequency terminal band, and the weight of the sub-band SNR is greater than other
  • the weight of the subband SNR of the subband, the enhanced segmentation signal to noise ratio SSNR is determined according to the weight of the subband SNR of each subband in the audio signal and the subband SNR of each subband, wherein the enhanced SSNR is greater than the reference SSNR.
  • the third determining unit 603 is configured to compare the enhanced SSNR with a voice activity detection VAD decision threshold to determine whether the audio signal is an active signal.
  • the apparatus 600 shown in FIG. 6 can determine the characteristics of the input audio signal, determine the enhanced SSNR in a corresponding manner according to the characteristics of the audio signal, and compare the enhanced SSNR with the VAD decision threshold, so that the active signal can be missed. The ratio is reduced.
  • the first determining unit 601 is specifically configured to determine, according to the sub-band signal to noise ratio SNR of the audio signal, the audio signal as the to-be-determined audio signal.
  • the first determining unit 601 is specifically configured to use the audio signal.
  • the audio signal is determined to be the audio signal to be determined.
  • the first determining unit 601 is configured to: in the audio signal, the number of the high frequency terminal strips in which the subband SNR is greater than the first preset threshold is greater than the second quantity and the audio signal neutron In the case where the number of low frequency terminal strips having an SNR smaller than the second preset threshold is greater than the third number, the audio signal is determined to be an audio signal to be determined.
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the first quantity, the second quantity, and the third quantity are also obtained based on statistics.
  • the first quantity as an example, in a large number of noise-containing voice unvoiced sample frames, the number of sub-bands whose SNR of the high-frequency terminal strip is larger than the first preset threshold is determined, and the first quantity is determined therefrom, so that the voice unvoiced samples are made.
  • the majority of the subband SNR in the frame is greater than the first preset threshold.
  • the number of high frequency terminal strips is greater than the first number.
  • the method of obtaining the second quantity is similar to the method of obtaining the first quantity.
  • the second number may be the same as the first quantity, and the second quantity may also be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is counted less than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice unvoiced samples The majority of the subband SNR in the frame is less than the second preset threshold.
  • the number of low frequency terminal strips is greater than the third number.
  • FIG. 7 is a structural block diagram of an apparatus according to an embodiment of the present invention.
  • the apparatus shown in Figure 7 is capable of performing the various steps of Figure 1 or Figure 2.
  • device 700 includes a processor 701 and a memory 702.
  • the processor 701 can be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a Field Programmable Gate Array (FPGA), or the like. Programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • the methods, steps, and logical block diagrams disclosed in the embodiments of the present invention may be implemented or carried out.
  • the general purpose processor may be a microprocessor or the processor or any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present invention may be directly implemented by the hardware decoding processor, or may be performed by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in random access memory Memory (Random Access Memory, RAM), Flash memory, Read-Only Memory (ROM), Programmable Read Only Memory, or Electrically Erasable Programmable Memory, Register, etc., are well-known storage media.
  • RAM Random Access Memory
  • ROM Read-Only Memory
  • Programmable Read Only Memory or Electrically Erasable Programmable Memory, Register, etc.
  • the storage medium is located in memory 702, and processor 701 reads the instructions in memory 702 and, in conjunction with its hardware, performs the steps of the above method.
  • the processor 701 is configured to determine that the input audio signal is an audio signal to be determined.
  • the processor 701 is configured to determine an enhanced segmentation signal to noise ratio SSNR of the audio signal, where the enhanced SSNR is greater than a reference SSNR.
  • the processor 701 is configured to compare the enhanced SSNR with a voice activity detection VAD decision threshold to determine whether the audio signal is an active signal.
  • the apparatus 700 shown in FIG. 7 can determine the characteristics of the input audio signal, determine the enhanced SSNR in a corresponding manner according to the characteristics of the audio signal, and compare the enhanced SSNR with the VAD decision threshold, so that the active signal can be missed. The ratio is reduced.
  • the processor 701 is specifically configured to determine, according to the subband SNR of the audio signal, the audio signal as an audio signal to be determined.
  • the processor 701 determines, according to the subband SNR of the audio signal, the audio signal is an audio signal to be determined
  • the processor 701 is specifically configured to: in the audio signal, the subband SNR is greater than When the number of the high frequency terminal strips of the first preset threshold is greater than the first number, the audio signal is determined to be the audio signal to be determined.
  • the processor 701 determines that the audio signal is an audio signal to be determined according to a subband SNR of the audio signal
  • the processor 701 is specifically configured to use a subband SNR in the audio signal. If the number of high frequency terminal strips greater than the first preset threshold is greater than the second number and the number of low frequency terminal strips in the audio signal where the subband SNR is less than the second preset threshold is greater than the third number, determining the audio signal is to be determined. Determine the audio signal.
  • the processor 701 determines that the audio signal is an audio signal to be determined according to a subband SNR of the audio signal
  • the processor 701 is specifically configured to use a subband in the audio signal.
  • the audio signal is determined to be the audio signal to be determined.
  • the processor 701 is specifically configured to determine, in the case that the audio signal is an unvoiced signal, the audio signal as an audio signal to be determined.
  • the audio signal is an unvoiced signal.
  • ZCR Zero-Crossing Rate
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the third preset threshold is also obtained based on statistics. Specifically, the third preset threshold is determined from the sub-band SNR of the large number of noise signals such that the sub-band SNR of most of the sub-bands of the noise signals is less than the value.
  • the first quantity, the second quantity, the third quantity, and the fourth quantity are also obtained based on statistics. Taking the first quantity as an example, in a large number of voice samples containing noise, the number of subbands whose SNR of the high frequency terminal strip is larger than the first preset threshold is determined, and the first quantity is determined therefrom, so that the voice samples are absolutely large. A majority of the high frequency terminal strips having a SNR greater than the first predetermined threshold are greater than the first number.
  • the method of determining the second quantity is similar to the method of determining the first quantity. The second number may be the same as the first quantity or may be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is calculated to be greater than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice samples are absolutely large.
  • a majority of the low frequency terminal band SNR greater than the second predetermined threshold is greater than the third number.
  • the statistical sub-band SNR is greater than the number of sub-bands of the third preset threshold, and the fourth quantity is determined therefrom, so that the majority of the voice samples are greater than the third preset.
  • the number of subband SNRs of the threshold is greater than the fourth number.
  • the processor 701 is specifically configured to determine a weight of a subband SNR of each subband in the audio signal, where the subband SNR is greater than the first preset threshold, and the weight of the high frequency terminal strip is greater than the subband SNR of other subbands. The weight is determined based on the weight of the subband SNR of each subband in the audio signal and the SNR of each subband.
  • the processor 701 is specifically configured to determine a reference SSNR of the audio signal, and determine an enhanced SSNR according to a reference SSNR of the audio signal.
  • the reference SSNR can be the SSNR calculated using Equation 1.1.
  • Benchmark SSNR In the calculation, the sub-band SNRs of the respective sub-bands included in the SSNR have the same weight in the SSNR.
  • the processor 701 is specifically configured to determine the enhanced SSNR by using the following formula:
  • SSNR represents the reference SSNR
  • SSNR' represents the enhanced SSNR
  • x and y represent enhancement parameters.
  • the value of x can be 1.07
  • the value of y can be 1.
  • the values of x and y may also be other suitable values such that the enhanced SSNR is properly greater than the reference SSNR.
  • the processor 701 is specifically configured to determine the enhanced SSNR by using the following formula:
  • SSNR represents the reference SSNR
  • SSNR' represents the enhanced SSNR
  • f(x), h(y) represents an enhancement function.
  • f(x) and h(y) may be functions related to the long-term SNR (LSNR) of the audio signal, and the long-term signal-to-noise ratio of the audio signal is for a long period of time.
  • Average SNR or weighted SNR For example, when lsnr is greater than 20, f(lsnr) may be equal to 1.1 and y(lsnr) may be equal to 2.
  • f(lsnr) When lsnr is less than 20 and greater than 17, f(lsnr) may be equal to 1.07, and y(lsnr) may be equal to 1. When lsnr is less than 17, f(lsnr) may be equal to 1, and y(lsnr) may be equal to zero.
  • f(x) and h(y) may also be in other suitable forms such that the enhanced SSNR is properly greater than the reference SSNR.
  • the processor 701 is specifically configured to compare the enhanced SSNR with a voice activity detection VAD decision threshold, and determine, according to the comparison structure, whether the audio signal is an active signal. Specifically, if the enhanced SSNR is greater than the VAD decision threshold, it is determined that the audio signal is an active signal. If the enhanced SSNR is less than the VAD decision threshold, then the audio signal is determined to be an inactive signal.
  • the reduced VAD decision threshold obtained after the reference VAD decision threshold is reduced may also be used by using a preset algorithm, and the reduced VAD decision threshold is used to determine whether the audio signal is an active signal.
  • the processor 701 can also be configured to reduce the VAD decision threshold by using a preset algorithm to obtain a reduced VAD decision threshold.
  • the processor 701 is specifically configured to compare the enhanced SSNR with the reduced VAD decision threshold to determine whether the audio signal is an active signal.
  • FIG. 8 is a structural block diagram of another apparatus according to an embodiment of the present invention.
  • the apparatus shown in Figure 8 is capable of performing the various steps of Figure 3.
  • device 800 includes a processor 801 and a memory 802.
  • the processor 801 can be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a Field Programmable Gate Array (FPGA), or the like. Programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • the methods, steps, and logical block diagrams disclosed in the embodiments of the present invention may be implemented or carried out.
  • the general purpose processor may be a microprocessor or the processor or any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present invention may be directly implemented by the hardware decoding processor, or may be performed by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read only memory or an electrically erasable programmable memory, a register, etc.
  • RAM random access memory
  • ROM read-only memory
  • programmable read only memory or an electrically erasable programmable memory
  • register etc.
  • the storage medium is located in memory 802, and processor 801 reads the instructions in memory 802 and, in conjunction with its hardware, performs the steps of the above method.
  • the processor 801 is configured to determine the input audio signal as the audio signal to be determined.
  • the processor 801 is configured to determine a weight of a sub-band signal-to-noise ratio SNR of each subband in the audio signal, where the subband SNR of the subband SNR greater than the first preset threshold has a weight greater than that of the other subbands
  • the weight of the subband SNR is determined according to the weight of the subband SNR of each subband in the audio signal and the subband SNR of each subband, wherein the enhanced SSNR is greater than the reference SSNR.
  • the processor 801 is configured to compare the enhanced SSNR with a voice activity detection VAD decision threshold to determine whether the audio signal is an active signal.
  • the apparatus 800 shown in FIG. 8 can determine the characteristics of the input audio signal, determine the enhanced SSNR in a corresponding manner according to the characteristics of the audio signal, and compare the enhanced SSNR with the VAD decision threshold, so that the active signal can be missed. The ratio is reduced.
  • the processor 801 is specifically configured to determine, according to the sub-band signal to noise ratio SNR of the audio signal, the audio signal as an audio signal to be determined.
  • the processor 801 is specifically configured to determine, when the number of the high frequency terminal strips in the audio signal that the subband signal to noise ratio SNR is greater than the first preset threshold is greater than the first quantity.
  • the audio signal is the audio signal to be judged.
  • the processor 801 is specifically configured to: in the audio signal, the number of the high frequency terminal strips in which the subband SNR is greater than the first preset threshold is greater than the second quantity and the subband SNR in the audio signal When the number of low frequency terminal strips smaller than the second preset threshold is greater than the third number, The audio signal is determined to be an audio signal to be determined.
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the first quantity, the second quantity, and the third quantity are also obtained based on statistics.
  • the first quantity as an example, in a large number of noise-containing voice unvoiced sample frames, the number of sub-bands whose SNR of the high-frequency terminal strip is larger than the first preset threshold is determined, and the first quantity is determined therefrom, so that the voice unvoiced samples are made.
  • the majority of the subband SNR in the frame is greater than the first preset threshold.
  • the number of high frequency terminal strips is greater than the first number.
  • the method of obtaining the second quantity is similar to the method of obtaining the first quantity.
  • the second number may be the same as the first quantity, and the second quantity may also be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is counted less than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice unvoiced samples The majority of the subband SNR in the frame is less than the second preset threshold.
  • the number of low frequency terminal strips is greater than the third number.
  • FIG. 9 is a structural block diagram of another apparatus according to an embodiment of the present invention.
  • the apparatus 900 shown in FIG. 9 can perform the various steps of FIG.
  • the apparatus 900 includes a first determining unit 901, a second determining unit 902, a third determining unit 903, and a fourth determining unit 904.
  • the first determining unit 901 is configured to determine that the input audio signal is an audio signal to be determined.
  • the second determining unit 902 is configured to acquire a reference SSNR of the audio signal.
  • the reference SSNR may be the SSNR calculated using Equation 1.1.
  • the third determining unit 903 is configured to reduce the reference VAD decision threshold by using a preset algorithm to obtain a reduced VAD decision threshold.
  • the reference VAD decision threshold may be a default VAD decision threshold, which may be pre-stored or temporarily calculated, wherein the calculation of the reference VAD decision threshold may be performed by using a prior art technique.
  • the preset algorithm may be to multiply the reference VAD decision threshold by a coefficient less than one, and other algorithms may be used.
  • the embodiment of the present invention does not limit the specific algorithm used. .
  • the preset algorithm may appropriately reduce the VAD decision threshold such that the enhanced SSNR is greater than the reduced VAD decision threshold, so that the proportion of the active signal being missed may be reduced.
  • the fourth determining unit 904 is configured to compare the reference SSNR with the reduced VAD decision threshold to determine whether the audio signal is an active signal.
  • the first determining unit 901 is specifically configured to determine, according to an SNR of the audio signal, the audio signal as an audio signal to be determined.
  • the first determining unit 901 determines that the audio signal is an audio signal to be determined according to an SNR of the audio signal
  • the first determining unit 901 is specifically configured to use a subband in the audio signal.
  • the audio signal is determined to be the audio signal to be determined.
  • the first determining unit 901 determines that the audio signal is an audio signal to be determined according to an SNR of the audio signal
  • the first determining unit 901 is specifically configured to use a subband in the audio signal. Determining that the audio signal is determined when the number of high frequency terminal strips whose SNR is greater than the first preset threshold is greater than the second number and the number of low frequency terminal strips in the audio signal where the subband SNR is less than the second preset threshold is greater than the third number The audio signal is to be judged.
  • the first determining unit 901 determines that the audio signal is an audio signal to be determined according to an SNR of the audio signal
  • the first determining unit 901 is specifically configured to use a neutron in the audio signal.
  • the audio signal is determined to be the audio signal to be determined.
  • the first determining unit 901 is specifically configured to determine, in the case that the audio signal is an unvoiced signal, the audio signal as an audio signal to be determined.
  • the audio signal is an unvoiced signal
  • whether the audio signal is an unvoiced signal can be determined by detecting a Zero-Crossing Rate (ZCR) of the audio signal.
  • ZCR Zero-Crossing Rate
  • the audio signal is determined to be an unvoiced signal, wherein the ZCR threshold is determined by a large number of experiments.
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the third preset threshold is also obtained based on statistics. Specifically, from a large number of noise signals The third preset threshold is determined in the subband SNR such that the subband SNR of most of the subbands of the noise signals is less than the value.
  • the first quantity, the second quantity, the third quantity, and the fourth quantity are also obtained based on statistics. Taking the first quantity as an example, in a large number of voice samples containing noise, the number of subbands whose SNR of the high frequency terminal strip is larger than the first preset threshold is determined, and the first quantity is determined therefrom, so that the voice samples are absolutely large. A majority of the high frequency terminal strips having a SNR greater than the first predetermined threshold are greater than the first number.
  • the method of determining the second quantity is similar to the method of determining the first quantity. The second number may be the same as the first quantity or may be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is calculated to be greater than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice samples are absolutely large.
  • a majority of the low frequency terminal band SNR greater than the second predetermined threshold is greater than the third number.
  • the statistical sub-band SNR is greater than the number of sub-bands of the third preset threshold, and the fourth quantity is determined therefrom, so that the majority of the voice samples are greater than the third preset.
  • the number of subband SNRs of the threshold is greater than the fourth number.
  • the apparatus 900 shown in FIG. 9 can determine the characteristics of the input audio signal, reduce the reference VAD decision threshold according to the characteristics of the audio signal, and compare the SSNR with the reduced VAD decision threshold, so that the active signal can be leaked.
  • the inspection ratio is reduced.
  • FIG. 10 is a structural block diagram of another apparatus according to an embodiment of the present invention.
  • the apparatus 1000 shown in FIG. 10 can perform the various steps of FIG.
  • the apparatus 1000 includes a processor 1001 and a memory 1002.
  • the processor 1001 may be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a Field Programmable Gate Array (FPGA), or the like. Programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • the methods, steps, and logical block diagrams disclosed in the embodiments of the present invention may be implemented or carried out.
  • the general purpose processor may be a microprocessor or the processor or any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present invention may be directly implemented by the hardware decoding processor, or may be performed by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read only memory or an electrically erasable programmable memory, a register, etc.
  • RAM random access memory
  • ROM read-only memory
  • programmable read only memory or an electrically erasable programmable memory
  • register etc.
  • the storage medium is located in the memory 1002, and the processor 1001 reads the instructions in the memory 1002 and completes the steps of the above method in combination with its hardware.
  • the processor 1001 is configured to determine the input audio signal as an audio signal to be determined.
  • the processor 1001 is configured to acquire a reference SSNR of the audio signal.
  • the reference SSNR may be the SSNR calculated using Equation 1.1.
  • the processor 1001 is configured to reduce the reference VAD decision threshold by using a preset algorithm to obtain a reduced VAD decision threshold.
  • the reference VAD decision threshold may be a default VAD decision threshold, which may be pre-stored or temporarily calculated, wherein the calculation of the reference VAD decision threshold may be performed by using a prior art technique.
  • the preset algorithm may be to multiply the reference VAD decision threshold by a coefficient less than one, and other algorithms may be used.
  • the embodiment of the present invention does not limit the specific algorithm used. .
  • the preset algorithm may appropriately reduce the VAD decision threshold such that the enhanced SSNR is greater than the reduced VAD decision threshold, so that the proportion of the active signal being missed may be reduced.
  • the processor 1001 is configured to compare the reference SSNR with the reduced VAD decision threshold to determine whether the audio signal is an active signal.
  • the processor 1001 is specifically configured to determine, according to an SNR of the audio signal, the audio signal as an audio signal to be determined.
  • the processor 1001 determines that the audio signal is an audio signal to be determined according to an SNR of the audio signal
  • the processor 1001 is specifically configured to: in the audio signal, the sub-band SNR is greater than the first When the number of high frequency terminal strips of the preset threshold is greater than the first number, the audio signal is determined to be an audio signal to be determined.
  • the processor 1001 determines that the audio signal is an audio signal to be determined according to an SNR of the audio signal
  • the processor 1001 is specifically configured to: in the audio signal, the sub-band SNR is greater than the first When the number of high frequency terminal strips of the preset threshold is greater than the second number and the number of low frequency terminal strips in which the subband SNR of the audio signal is less than the second preset threshold is greater than the third quantity, determining the audio signal as the audio signal to be determined .
  • the processor 1001 determines that the audio signal is an audio signal to be determined according to an SNR of the audio signal
  • the processor 1001 is specifically configured to use a value of a neutron band SNR in the audio signal.
  • the audio signal is determined to be an audio signal to be determined.
  • the processor 1001 is specifically configured to determine, in the case that the audio signal is an unvoiced signal, the audio signal as an audio signal to be determined.
  • the technology in the field The operator can understand that there are a variety of methods for detecting whether an audio signal is an unvoiced signal.
  • whether the audio signal is an unvoiced signal can be determined by detecting a Zero-Crossing Rate (ZCR) of the audio signal.
  • ZCR Zero-Crossing Rate
  • the audio signal is determined to be an unvoiced signal, wherein the ZCR threshold is determined by a large number of experiments.
  • the first preset threshold and the second preset threshold may be obtained according to a large number of voice samples. Specifically, in a large number of voice unvoiced samples containing background noise, the subband SNR of the high frequency terminal strip is counted, and the first preset threshold is determined therefrom, so that the subband SNR of most of the high frequency terminal strips in the unvoiced samples Both are greater than the threshold. Similarly, the subband SNR of the low frequency terminal strip is counted in the speech unvoiced samples, and the second preset threshold is determined therefrom such that the subband SNR of most of the low frequency terminal strips of the speech unvoiced samples is less than the threshold.
  • the third preset threshold is also obtained based on statistics. Specifically, the third preset threshold is determined from the sub-band SNR of the large number of noise signals such that the sub-band SNR of most of the sub-bands of the noise signals is less than the value.
  • the first quantity, the second quantity, the third quantity, and the fourth quantity are also obtained based on statistics. Taking the first quantity as an example, in a large number of voice samples containing noise, the number of subbands whose SNR of the high frequency terminal strip is larger than the first preset threshold is determined, and the first quantity is determined therefrom, so that the voice samples are absolutely large. A majority of the high frequency terminal strips having a SNR greater than the first predetermined threshold are greater than the first number.
  • the method of determining the second quantity is similar to the method of determining the first quantity. The second number may be the same as the first quantity or may be different from the first quantity.
  • the sub-band SNR of the low-frequency terminal strip is calculated to be greater than the number of sub-bands of the second preset threshold, and the third quantity is determined therefrom, so that the voice samples are absolutely large.
  • a majority of the low frequency terminal band SNR greater than the second predetermined threshold is greater than the third number.
  • the statistical sub-band SNR is greater than the number of sub-bands of the third preset threshold, and the fourth quantity is determined therefrom, so that the majority of the voice samples are greater than the third preset.
  • the number of subband SNRs of the threshold is greater than the fourth number.
  • the apparatus 1000 shown in FIG. 10 can determine the characteristics of the input audio signal, reduce the reference VAD decision threshold according to the characteristics of the audio signal, and compare the SSNR with the reduced VAD decision threshold, so that the active signal can be leaked.
  • the inspection ratio is reduced.
  • the disclosed systems, devices, and methods may be implemented in other manners.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the functions may be stored in a computer readable storage medium if implemented in the form of a software functional unit and sold or used as a standalone product.
  • the technical solution of the present invention which is essential or contributes to the prior art, or a part of the technical solution, may be embodied in the form of a software product, which is stored in a storage medium, including
  • the instructions are used to cause a computer device (which may be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the methods described in various embodiments of the present invention.
  • the foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Computational Linguistics (AREA)
  • Acoustics & Sound (AREA)
  • Signal Processing (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Telephone Function (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)
  • Noise Elimination (AREA)
  • Telephonic Communication Services (AREA)
  • Signal Processing Not Specific To The Method Of Recording And Reproducing (AREA)
  • Circuit For Audible Band Transducer (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

一种检测音频信号的方法和装置,包括:确定输入的音频信号为待判断音频信号(101);确定该音频信号的增强分段信噪比SSNR(102),其中该增强SSNR大于基准SSNR;将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号(103)。所提供的方法和装置能够准确地分辨活动语音和非活动语音。

Description

检测音频信号的方法和装置
本申请要求于2014年3月12日提交中国专利局、申请号为201410090386.X、发明名称为“检测音频信号的方法和装置”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本发明实施例涉及信号处理技术领域,并且更具体地,涉及检测音频信号的方法和装置。
背景技术
语音活动检测(Voice Activity Detection,VAD)是一种广泛应用与语音通信、人机交互等领域的关键技术,VAD也可以被称为声音活动检测(Sound Activity Detection,SAD)。它的作用是检测输入的音频信号中是否有活动性信号,其中活动性信号是相对于非活动信号而言(例如环境背景噪音、静音等)。典型的活动信号包括语音、音乐等。VAD的原理是从输入的音频信号中提取一个或多个特征参数,根据这一个或多个特征参数确定一个或多个特征值,然后将这一个或多个特征值与一个或多个门限值进行比较。
现有技术中的基于分段信噪比(Segmental Signal to Noise Ratio,SSNR)的活动信号检测方法是将输入的音频信号在频带上划分为多个子带信号,计算该音频信号在每一个子带的能量,通过将该音频信号在每一个子带的能量与一个估计出的背景噪声信号在每个子带的能量做对比,获得该音频信号在每个子带上的信噪比(Signal-to-Noise Ratio,SNR)。然后根据每个子带上的子带SNR确定SSNR,将SSNR与预设的VAD判决门限进行比较,如果该SSNR超过该VAD判决门限,则该音频信号为活动信号;如果该SSNR没有超过该VAD判决门限,则该音频信号为非活动信号。
典型的一种计算SSNR的方法是将该音频信号所有子带SNR相加,得到的结果就是SSNR。例如,可以采用公式1.1确定SSNR:
Figure PCTCN2014092694-appb-000001
...........................................................公式1.1
其中,k表示第k个子带,snr(k)表示第k个子带的子带SNR,N表示该音频信号总共被划分为子带的子带个数。
通过上述计算SSNR的方法检测活动语音时,可能会造成活动语音的漏检。
发明内容
本发明实施例提供了检测音频信号的方法和装置,能够准确地分辨活动语音和非活动语音。
第一方面,本发明实施例提供一种检测音频信号的方法,该方法包括:确定输入的音频信号为待判断音频信号;确定该音频信号的增强分段信噪比SSNR,其中该增强SSNR大于基准SSNR;将该增强SSNR与语音活动检测VAD判决门限进行比较,确定该音频信号是否为活动信号。
结合第一方面,在第一方面的第一种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:根据该音频信号的子带信噪比SNR,确定该音频信号为待判断音频信号。
结合第一方面的第一种可能的实现方式,在第一方面的第二种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
结合第一方面的第一种可能的实现方式,在第一方面的第三种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
结合第一方面的第一种可能的实现方式,在第一方面的第四种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
结合第一方面,在第一方面的第五种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。
结合第一方面的第二种可能的实现方式或第三种可能的实现方式,在第一方面的第六种可能的实现方式中,该确定该音频信号的增强分段信噪比 SSNR,包括:确定该音频信号中各个子带的子带SNR的权重,其中该子带SNR大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重;根据该音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定该增强SSNR。
结合第一方面或第一方面的第一种可能的实现方式至第一方面的第一方面的第五种可能的实现方式中的任一种可能的实现方式,在第一方面的第七种可能的实现方式中,该确定该音频信号的增强分段信噪比SSNR,包括:确定该音频信号的基准SSNR;根据该音频信号的基准SSNR,确定增强SSNR。
结合第一方面的第七种可能的实现方式,在第一方面的第八种可能的实现方式中,该根据该音频信号的基准SSNR,确定增强SSNR,包括:使用以下公式确定该增强SSNR:SSNR′=x*SSNR+y,其中,SSNR表示该基准SSNR,SSNR′表示该增强SSNR,x和y表示增强参数。
结合第一方面的第七种可能的实现方式,在第一方面的第九种可能的实现方式中,该根据该音频信号的基准SSNR,确定增强SSNR,包括:使用以下公式确定该增强SSNR:SSNR′=f(x)*SSNR+h(y),其中,SSNR表示该基准SSNR,SSNR′表示该增强SSNR,f(x)、h(y)表示增强函数。
结合第一方面或第一方面的上述任一种可能的实现方式,在第一方面的第十种可能的实现方式中,该将该增强SSNR与语音活动检测VAD判决门限进行比较前进一步包括:使用预置算法减小该VAD判决门限,获得减小后的VAD判决门限;该将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号具体包括:将该增强SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
第二方面,本发明实施例提供一种检测音频信号的方法,该方法包括:确定输入的音频信号为待判断音频信号;确定该音频信号中各个子带的子带信噪比SNR的权重,其中该子带SNR大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重;根据该音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定增强分段信噪比SSNR,其中该增强SSNR大于基准SSNR;将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号。
结合第二方面,在第二方面的第一种可能的实现方式中,该确定输入的 音频信号为待判断音频信号,包括:根据该音频信号的子带SNR,确定该音频信号为待判断音频信号。
结合第二方面的第一种可能的实现方式,在第二方面的第二种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
结合第二方面的第一种可能的实现方式,在第二方面的第三种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
第三方面,本发明实施例提供一种检测音频信号的方法,该方法包括:确定输入的音频信号为待判断音频信号;获取该音频信号的基准分段信噪比SSNR;使用预置算法减小基准语音活动检测VAD判决门限,获得减小后的VAD判决门限;将该基准SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
结合第三方面,在第三方面的第一种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:根据该音频信号的子带信噪比SNR,确定该音频信号为待判断音频信号。
结合第三方面的第一种可能的实现方式,在第三方面的第二种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
结合第三方面的第一种可能的实现方式,在第三方面的第三种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
结合第三方面的第一种可能的实现方式,在第三方面的第四种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下, 确定该音频信号为待判断音频信号。
结合第三方面,在第三方面的第五种可能的实现方式中,该确定输入的音频信号为待判断音频信号,包括:在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。
第四方面,本发明实施例提供一种装置,该装置包括:第一确定单元,用于确定输入的音频信号为待判断音频信号;第二确定单元,用于确定该音频信号的增强分段信噪比SSNR,其中该增强SSNR大于基准SSNR;第三确定单元,用于将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号。
结合第四方面,在第四方面的第一种可能的实现方式中,该第一确定单元,具体用于根据该音频信号的子带信噪比SNR,确定该音频信号为待判断音频信号。
结合第四方面的第一种可能的实现方式,在第四方面的第二种可能的实现方式中,该第一确定单元,具体用于在该音频信号中子带信噪比SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
结合第四方面的第一种可能的实现方式,在第四方面的第三种可能的实现方式中,该第一确定单元,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
结合第四方面的第一种可能的实现方式,在第四方面的第四种可能的实现方式中,该第一确定单元,具体用于在该音频信号中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
结合第四方面,在第四方面的第五种可能的实现方式中,该第一确定单元,具体用于在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。
结合第四方面的第二种可能的实现方式或第四方面的第三种可能的实现方式,在第四方面的第六种可能的实现方式中,该第二确定单元,具体用于确定该音频信号中各个子带的子带SNR的权重,其中该子带SNR大于第 一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重,根据该音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定该增强SSNR。
结合第四方面或第四方面的第一种可能的实现方式至第四方面的第五种可能的实现方式中的任一种可能的实现方式,在第四方面的第七种可能的实现方式中,第二确定单元,具体用于确定该音频信号的基准SSNR,根据该音频信号的基准SSNR,确定该增强SSNR。
结合第四方面的第七种可能的实现方式,在第四方面的第八种可能的实现方式中,该第二确定单元,具体用于使用以下公式确定该增强SSNR:SSNR′=x*SSNR+y,其中,SSNR表示该基准SSNR,SSNR′表示该增强SSNR,x和y表示增强参数。
结合第四方面的第七种可能的实现方式,在第四方面的第九种可能的实现方式中,该第二确定单元,具体用于使用以下公式确定该增强SSNR:SSNR′=f(x)*SSNR+h(y),其中,SSNR表示该基准SSNR,SSNR′表示该增强SSNR,f(x)、h(y)表示增强函数。
结合第四方面或第四方面的上述任一种可能的实现方式,在第四方面的第十种可能的实现方式中,该装置还包括第四确定单元;该第四确定单元,用于使用预置算法减小该VAD判决门限,获得减小后的VAD判决门限;该第三确定单元,具体用于将该增强SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
第五方面,本发明实施例提供一种装置,该装置包括:第一确定单元,用于确定输入的音频信号为待判断音频信号;第二确定单元,用于确定该音频信号中各个子带的子带信噪比SNR的权重,其中该子带SNR大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重,根据该音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定增强分段信噪比SSNR,其中该增强SSNR大于基准SSNR;第三确定单元,用于将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号。
结合第五方面,在第五方面的第一种可能的实现方式中,该第一确定单元,具体用于根据该音频信号的子带信噪比SNR,确定该音频信号为待判断音频信号。
结合第五方面的第一种可能的实现方式,在第五方面的第二种可能的实现方式中,该第一确定单元,具体用于在该音频信号中子带信噪比SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
结合第五方面的第一种可能的实现方式,在第五方面的第三种可能的实现方式中,该第一确定单元,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
第六方面,本发明实施例提供一种装置,该装置包括:第一确定单元,用于确定输入的音频信号为待判断音频信号;第二确定单元,用于获取该音频信号的基准分段信噪比SSNR;第三确定单元,用于使用预置算法减小基准语音活动检测VAD判决门限,获得减小后的VAD判决门限;第四确定单元,用于将该基准SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
结合第六方面,在第六方面的第一种可能的实现方式中,该第一确定单元,具体用于根据该音频信号的子带信噪比SNR,确定该音频信号为待判断音频信号。
结合第六方面的第一种可能的实现方式,在第六方面的第二种可能的实现方式中,该第一确定单元,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
结合第六方面的第一种可能的实现方式,在第六方面的第三种可能的实现方式中,该第一确定单元,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
结合第六方面的第一种可能的实现方式,在第六方面的第四种可能的实现方式中,该第一确定单元,具体用于在该音频信号中中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
结合第六方面,在第六方面的第五种可能的实现方式中,该第一确定单元,具体用于在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。
根据本发明实施例所提供的方法,可以确定音频信号的特征,根据音频信号的特征,采用相应的方式确定增强SSNR,并采用该增强SSNR与VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低
附图说明
为了更清楚地说明本发明实施例的技术方案,下面将对本发明实施例中所需要使用的附图作简单地介绍,显而易见地,下面所描述的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1是根据本发明实施例提供的检测音频信号的方法的示意性流程图。
图2是根据本发明实施例提供的检测音频信号的方法的示意性流程图。
图3是根据本发明实施例提供的检测音频信号的方法的示意性流程图。
图4是根据本发明实施例提供的检测音频信号的方法的示意性流程图。
图5是根据本发明实施例提供的装置的结构框图。
图6是根据本发明实施例提供的另一装置的结构框图。
图7是根据本发明实施例提供的装置的结构框图。
图8是根据本发明实施例提供的另一装置的结构框图。
图9是根据本发明实施例提供的另一装置的结构框图。
图10是根据本发明实施例提供的另一装置的结构框图。
具体实施方式
下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所述的实施例是本发明的一部分实施例,而不是全部实施例。基于本发明中的实施例,本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例,都应属于本发明保护的范围。
图1是根据本发明实施例提供的检测音频信号的方法的示意性流程图。
101,确定输入的音频信号为待判断音频信号。
102,确定该音频信号的增强SSNR,其中该增强SSNR大于基准SSNR。
103,将该增强SSNR与VAD判决门限比较,确定该音频信号是否为活动信号。
在本发明的实施例中,在将增强SSNR与VAD判决门限进行比较时,可以使用基准VAD判决门限,也可以使用预置算法减小基准VAD判决门限后获得的减小后的VAD判决门限。其中,基准VAD判决门限可以是默认的VAD判决门限,该基准VAD判决门限可以是预先存储的,也可以是临时计算获得,其中基准VAD判决门限的计算可以采用现有公知技术。在使用预置算法减小基准VAD判决门限时,该预置算法可以是将基准VAD判决门限乘于一个小于1的系数,也可以采用其他算法,本发明实施例并不限定所采用的具体算法。
在采用传统的SSNR计算方法计算一些音频信号的SSNR时,这些音频信号的SSNR可能低于预设的VAD判决门限。但是,实际上这些音频信号是活动音频信号。这是由于这些音频信号的特性导致的。例如,在环境SNR较低的情况下,高频部分的子带SNR会显著降低。并且,由于通常会采用心理声学理论划分子带,高频部分的子带SNR对SSNR的贡献较低。在此情况下,对一些能量主要集中在相对高频部分的信号,如清音信号,采用传统的SSNR计算方法计算出的SSNR可能低于VAD判决门限,这就造成活动信号的漏检。又如,一些音频信号中,音频信号的能量较平坦的分布在频谱上,但是该音频信号的整体能量较低。这样,在环境SNR较低的情况下,采用传统的SSNR计算方法计算出的SSNR也可能低于VAD判决门限。图1所示的方法通过适当的提高SSNR的方式,使得SSNR可以大于VAD判决门限,从而能够有效地降低活动信号漏减的比例。
图2是根据本发明实施例提供的检测音频信号的方法的示意性流程图。
201,确定输入的音频信号的子带SNR。
将输入音频信号的频谱划分为N个子带,其中N为大于1的正整数。具体地,可以采用心理声学理论对该音频信号的频谱进行划分。在采用心理声学理论划分音频信号的频谱的情况下,越靠近低频的子带宽度越窄,越靠近高频的子带宽度越宽。当然,也可以采用其他的方式划分该音频信号的频谱,例如将该音频信号的频谱等分为N个子带等方式。计算输入音频信号每个子带的子带SNR,其中该子带SNR为该子带的能量与背景噪声在该子带上的能量之比。背景噪声的子带能量一般是通过背景噪声估计器估计出来的 估计值。如何采用背景噪声估计器估计出每个子带对应的背景噪声能量是本领域的公知技术,因此,这里就不必赘述。本领域技术人员可以理解,该子带SNR可以是直接的能量比值,也可以是直接能量比值的其他表现形式,例如对数子带SNR。此外,本领域技术人员还可以理解,该子带SNR还可以是对直接子带SNR做线性或非线性处理后的子带SNR或者其他的变形。以下公式是子带SNR的直接能量比值:
snr(k)=E(k)/En(k),......................................................公式1.2
其中,snr(k)表示第k子带的子带SNR,E(k)和En(k)分别表示第k子带的能量和背景噪声在第k子带上的能量。对数子带SNR可以表示为:snrlog(k)=10×log10snr(k),其中snrlog(k)表示第k子带的对数子带SNR,snr(k)表示采用公式1.2计算出的第k子带的子带SNR。本领域技术人员还可以理解,用于计算子带SNR的子带能量既可以是输入音频信号在子带上的能量,也可以是输入音频信号在子带上的能量去除背景噪声在该子带上的能量之后的能量。SNR的计算只要不脱离SNR的意义即可。
202,确定输入的音频信号为待判断音频信号。
可选的,作为一个实施例,该确定输入的音频信号为待判断音频信号包括:可以是根据步骤201中确定的该音频信号的子带SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,在根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量的情况下,确定该音频信号为待判断音频信号。在本发明实施例中,一帧音频信号的高频端和低频端是相对而言的,即频率相对高一些的部分为高频端,频率相对低一些的部分为低频端。
可选的,作为另一个实施例,在根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,该确定输入的音频信号为待判断音频信 号,包括:在该音频信号中的子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
该第三预设门限也是根据统计得到的。具体来说,从大量的噪声信号的子带SNR中确定第三预设门限,使得这些噪声信号中的绝大多数子带的子带SNR都小于该值。
第一数量、第二数量、第三数量和第四数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音清音样本帧中,统计高频端子带的子带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音清音样本帧中绝大多数的子带SNR大于第一预设门限的高频端子带的数量大于该第一数量。获取第二数量的方法与获取第一数量的方法类似。第二数量可以与第一数量相同,第二数量也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音清音样本帧中,统计低频端子带的子带SNR小于第二预设门限的子带数量,从中确定第三数量,使得这些语音清音样本帧中绝大多数的子带SNR小于第二预设门限的低频端子带的数量大于该第三数量。对于第四数量,在大量的噪声信号帧中,统计子带SNR小于第三预设门限的子带数量,从中确定第四数量,使得这些噪声样本帧中绝大多数的子带SNR小于第三预设门限的子带的数量大于该第四数量。
可选的,作为另一个实施例,可以通过判断输入的音频信号是否为清音信号来确定输入的音频信号是否为待判断音频信号。在此情况下,判断该音频信号是否为待判断音频信号时不需要确定该音频信号的子带SNR。换句话说,在判断该音频信号是否为待判断音频信号时不需要执行步骤201。具体地,该确定输入的音频信号为待判断音频信号,包括:在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。具体地,本领域技术人员可以理解,可以有多种用于检测音频信号是否为清音信号的方法。例如,可以通过检测该音频信号的时域过零率(Zero-Crossing Rate,ZCR)来 确定该音频信号是否为清音信号。具体地,在该音频信号的ZCR大于ZCR阈值的情况下,确定该音频信号为清音信号,其中该ZCR阈值是通过大量实验确定的。
203,确定该音频信号的增强SSNR,其中该增强SSNR大于基准SSNR。
该基准SSNR可以是采用公式1.1计算出来的SSNR。从公式1.1可以看出,在计算基准SSNR时,没有对任何一个子带的子带SNR进行加权处理,也就是说,在计算基准SSNR时各个子带的子带SNR的权重相同。
可选的,作为一个实施例,在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,或者,在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且在该音频信号中子带SNR小于第二预设门限的低频端子带的数量小于第三数量的情况下,该确定该音频信号的增强SSNR,包括:确定该音频信号中各个子带的子带SNR的权重,其中该子带SNR大于第一预设门限的高频端子带的权重大于其他子带的子带SNR的权重,根据该音频信号中各个子带的子带SNR的权重和各个子带的子带SNR,确定该增强SSNR。
例如,如果将该音频信号按照心理声学理论划分为20个子带,即子带0至子带19。如果子带18和子带19均大于第一预设值T1,则可以增加四个子带,即子带20至子带23。具体来说,可以将信噪比大于T1的子带18划分为子带18a、子带18b和子带18c,子带19划分为子带19a、子带19b和子带19c。这样,子带18可以看作是子带18a、子带18b和子带18c的母子带,子带19可以看作是子带19a、子带19b和子带19c的母子带。子带18a、子带18b和子带18c的信噪比的取值与其母子带的信噪比取值相同,子带19a、子带19b和子带19c的信噪比的取值与其母子带的信噪比的取值相同。这样,就将原有划分的20个子带重新划分为24个子带。由于在进行活动信号检测时,VAD仍然是按照20个子带进行设计的,因此需要将24个子带映射回20个子带,来确定增强SSNR。综上,采用增加该子带SNR大于该第一预设门限的高频端子带的数量的方式来确定该增强SSNR时,可以采用以下公式进行计算:
Figure PCTCN2014092694-appb-000002
.................公式1.3
其中,SSNR′表示该增强SSNR。snr(k)表示第k子带的子带SNR。
如果采用公式1.1计算的SSNR为基准SSNR,则计算出来的基准SSNR为
Figure PCTCN2014092694-appb-000003
显然,对于第一类音频信号采用公式1.3计算出来的增强SSNR的值大于采用公式1.1计算出来的基准SSNR的值。
又如,如果将该音频信号按照心理声学理论划分为20个子带,即子带0至子带19。如果snr(18)和snr(19)均大于第一预设值T1,且snr(0)到snr(17)均小于第二预设值T2,则可以采用以下公式确定该增强SSNR:
Figure PCTCN2014092694-appb-000004
.......................公式1.4
其中,SSNR′表示该增强SSNR,snr(k)表示第k子带的子带SNR,a1和a2为增加权重参数并且a1和a2的取值使得a1×snr(18)+a2×snr(19)大于snr(18)+snr(19)。显然,采用公式1.4计算出来的增强SSNR的值大于采用公式1.1计算出来的基准SSNR的值。
可选的,作为另一实施例,该确定该音频信号的增强SSNR,包括:确定该音频信号的基准SSNR,根据该音频信号的基准SSNR,确定增强SSNR。
可选的,可以使用以下公式确定该增强SSNR:
SSNR′=x*SSNR+y,......................................................公式1.5
其中,SSNR表示该音频信号的基准SSNR,SSNR′表示该增强SSNR,x和y表示增强参数。例如,x的取值可以为1.05,y的取值可以为1。本领域技术人员可以理解,x和y的取值还可以是其他合适的值,使得增强SSNR恰当的大于基准SSNR。
可选的,可以使用以下公式确定该增强SSNR:
SSNR′=f(x)*SSNR+h(y),.............................................公式1.6
其中,SSNR表示该音频信号的原始SSNR,SSNR′表示该增强SSNR,f(x)、h(y)表示增强函数。例如,f(x)和h(y)可以是与该音频信号的长时信噪比(Long-term SNR,LSNR)相关的函数,音频信号的长时信噪比为一段较长时间内的平均SNR或加权SNR。例如,当lsnr大于20时,f(lsnr)可以等于1.1,y(lsnr)可以等于2。当lsnr小于20且大于15时,f(lsnr)可以等于1.05,y(lsnr)可以等于1。当lsnr小于15时,f(lsnr)可以等于1,y(lsnr)可以等于0。本领域技术人员可以理解,f(x)和h(y)还可以是其他合适的形式,使得增强SSNR恰当的大于基准SSNR。
204,将该增强SSNR与VAD判决门限比较,确定该音频信号是否为活动信号。
具体来说,将该增强SSNR与VAD判决门限比较,如果该增强SSNR大于该VAD判决门限,则确定该音频信号为活动信号。否则确定该音频信号为非活动信号。
可选的,作为另一个实施例,在将该增强SSNR与VAD判决门限进行比较前,该方法还可以包括:使用预置算法减小该VAD判决门限,获得减小后的VAD判决门限。在此情况下,将该增强SSNR与VAD判决门限比较具体包括:将该增强SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。基准VAD判决门限可以是默认的VAD判决门限,该基准VAD判决门限可以是预先存储的,也可以是临时计算获得,其中基准VAD判决门限的计算可以采用现有公知技术。在使用预置算法减小基准VAD判决门限时,该预置算法可以是将基准VAD判决门限乘于一个小于1的系数,也可以采用其他算法,本发明实施例并不限定所采用的具体算法。该预置算法可以适当减小VAD判决门限,使得增强SSNR大于该减小后的VAD判决门限,从而可以使得活动信号被漏减的比例降低。
根据图2所示的方法,确定音频信号的特征,根据音频信号的特征,采用相应的方式确定增强SSNR,并采用该增强SSNR与VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低。
图3是根据本发明实施例提供的检测音频信号的方法的示意性流程图。
301,确定输入的音频信号为待判断音频信号。
302,确定该音频信号中各个子带的子带SNR的权重,其中该子带SNR大于第一预设门限的包频段子带的子带SNR的权重大于其他子带的子带SNR的权重。
303,根据该音频信号中各个子带的子带SNR的权重和各个子带的子带SNR,确定增强SSNR,其中该增强SSNR大于基准SSNR。
该基准SSNR可以是采用公式1.1计算出来的SSNR。从公式1.1可以看出,在计算基准SSNR时,没有对任何一个子带的子带SNR进行加权处理,也就是说,在计算基准SSNR时各个子带的子带SNR的权重相同。
例如,如果将该音频信号按照心理声学理论划分为20个子带,即子带0至子带19。如果子带18和子带19均大于第一预设值T1,则可以增加四个 子带,即子带20至子带23。具体来说,可以将信噪比大于T1的子带18划分为子带18a、子带18b和子带18c,子带19划分为子带19a、子带19b和子带19c。这样,子带18可以看作是子带18a、子带18b和子带18c的母子带,子带19可以看作是子带19a、子带19b和子带19c的母子带。子带18a、子带18b和子带18c的信噪比的取值与其母子带的信噪比取值相同,子带19a、子带19b和子带19c的信噪比的取值与其母子带的信噪比的取值相同。这样,就将原有划分的20个子带重新划分为24个子带。由于在进行活动信号检测时,VAD仍然是按照20个子带进行设计的,因此需要将24个子带映射回20个子带,来确定增强SSNR。综上,采用增加该子带SNR大于该第一预设门限的高频端子带的数量的方式来确定该增强SSNR时,可以采用以下公式进行计算:
Figure PCTCN2014092694-appb-000005
..................公式1.3
其中,SSNR′表示该增强SSNR。snr(k)表示第k子带的子带SNR。
如果采用公式1.1计算的SSNR为基准SSNR,则计算出来的基准SSNR为
Figure PCTCN2014092694-appb-000006
显然,对于第一类音频信号采用公式1.3计算出来的增强SSNR的值大于采用公式1.1计算出来的基准SSNR的值。
又如,如果将该音频信号按照心理声学理论划分为20个子带,即子带0至子带19。如果snr(18)和snr(19)均大于第一预设值T1,且snr(0)到snr(17)均小于第二预设值T2,则可以采用以下公式确定该增强SSNR:
Figure PCTCN2014092694-appb-000007
......................公式1.4
其中,SSNR′表示该增强SSNR,snr(k)表示第k子带的子带SNR,a1和a2为增加权重参数并且a1和a2的取值使得a1×snr(18)+a2×snr(19)大于snr(18)+snr(19)。显然,采用公式1.4计算出来的增强SSNR的值大于采用公式1.1计算出来的基准SSNR的值。
304,将该增强SSNR与VAD判决门限比较,确定该音频信号是否为活动信号。
具体来说,将该增强SSNR与VAD判决门限比较,如果该增强SSNR大于该VAD判决门限,则确定该音频信号为活动信号。否则确定该音频信 号为非活动信号。
图3所述的方法可以确定音频信号的特征,根据音频信号的特征,采用相应的方式确定增强SSNR,并采用该增强SSNR与VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低。
进一步,该确定输入的音频信号为待判断音频信号,包括,根据该音频信号的子带SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,该确定该音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,在根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,该确定该音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量的情况下,确定该音频信号为待判断音频信号。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
第一数量、第二数量和第三数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音清音样本帧中,统计高频端子带的子带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音清音样本帧中绝大多数的子带SNR大于第一预设门限的高频端子带的数量大于该第一数量。获取第二数量的方法与获取第一数量的方法类似。第二数量可以与第一数量相同,第二数量也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音清音样本帧中,统计低频端子带的子带SNR小于第二预设门限的子带数量,从中确定第三数量,使得这些语音清音样本帧中绝大多数的子带SNR小于第二预设门限的低频端子带的数量大于该第三数量。
图1至图3的实施例通过使用增强SSNR的方式判断输入的音频信号是 否为活动信号。图4所示的方法是通过减小VAD判决门限的方式判断输入的音频信号是否为活动信号。
图4是根据本发明实施例提供的检测音频信号的方法的示意性流程图。
401,确定输入的音频信号为待判断音频信号。
可选的,作为一个实施例,该确定输入的音频信号为待判断音频信号包括:可以是根据步骤201中确定的该音频信号的子带SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,在根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,在根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,该确定输入的音频信号为待判断音频信号,包括:在该音频信号中的子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
该第三预设门限也是根据统计得到的。具体来说,从大量的噪声信号的子带SNR中确定第三预设门限,使得这些噪声信号中的绝大多数子带的子带SNR都小于该值。
第一数量、第二数量、第三数量和第四数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音清音样本帧中,统计高频端子带的子 带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音清音样本帧中绝大多数的子带SNR大于第一预设门限的高频端子带的数量大于该第一数量。获取第二数量的方法与获取第一数量的方法类似。第二数量可以与第一数量相同,第二数量也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音清音样本帧中,统计低频端子带的子带SNR小于第二预设门限的子带数量,从中确定第三数量,使得这些语音清音样本帧中绝大多数的子带SNR小于第二预设门限的低频端子带的数量大于该第三数量。对于第四数量,在大量的噪声信号帧中,统计子带SNR小于第三预设门限的子带数量,从中确定第四数量,使得这些噪声样本帧中绝大多数的子带SNR小于第三预设门限的子带的数量大于该第四数量。
可选的,作为另一个实施例,可以通过判断输入的音频信号是否为清音信号来确定输入的音频信号是否为待判断音频信号。在此情况下,判断该音频信号是否为待判断音频信号时不需要确定该音频信号的子带SNR。换句话说,在判断该音频信号是否为待判断音频信号时不需要执行步骤201。具体地,该确定输入的音频信号为待判断音频信号,包括:在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。具体地,本领域技术人员可以理解,可以有多种用于检测音频信号是否为清音信号的方法。例如,可以通过检测该音频信号的时域过零率(Zero-Crossing Rate,ZCR)来确定该音频信号是否为清音信号。具体地,在该音频信号的ZCR大于ZCR阈值的情况下,确定该音频信号为清音信号,其中该ZCR阈值是通过大量实验确定的。
402,获取该音频信号的基准SSNR。
具体地,该基准SSNR可以是采用公式1.1计算出来的SSNR。
403,使用预置算法减小基准VAD判决门限,获得减小后的VAD判决门限。
具体地,基准VAD判决门限可以是默认的VAD判决门限,该基准VAD判决门限可以是预先存储的,也可以是临时计算获得,其中基准VAD判决门限的计算可以采用现有公知技术。在使用预置算法减小基准VAD判决门限时,该预置算法可以是将基准VAD判决门限乘于一个小于1的系数,也可以采用其他算法,本发明实施例并不限定所采用的具体算法。该预置算法可以适当减小VAD判决门限,使得增强SSNR大于该减小后的VAD判决门 限,从而可以使得活动信号被漏减的比例降低。
404,将该基准SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
在采用传统的SSNR计算方法计算一些音频信号的SSNR时,这些音频信号的SSNR可能低于预设的VAD判决门限。但是,实际上这些音频信号是活动音频信号。这是由于这些音频信号的特性导致的。例如,在环境SNR较低的情况下,高频部分的子带SNR会显著降低。并且,由于通常会采用心理声学理论划分子带,高频部分的子带SNR对SSNR的贡献较低。在此情况下,对一些能量主要集中在相对高频部分的信号,如清音信号,采用传统的SSNR计算方法计算出的SSNR可能低于VAD判决门限,这就造成活动信号的漏检。又如,一些音频信号中,音频信号的能量较平坦的分布在频谱上,但是该音频信号的整体能量较低。这样,在环境SNR较低的情况下,采用传统的SSNR计算方法计算出的SSNR也可能低于VAD判决门限。图4所示的方法通过降低VAD判决门限的方式,使得采用传统的SSNR计算方法计算出的SSNR大于VAD判决门限,从而能够有效地降低活动信号漏减的比例。
图5是根据本发明实施例提供的装置的结构框图。图5所示的装置能够执行图1或图2的各个步骤。如图5所示,装置500包括第一确定单元501、第二确定单元502和第三确定单元503。
第一确定单元501,用于确定输入的音频信号为待判断音频信号。
第二确定单元502,用于确定该音频信号的增强分段信噪比SSNR,其中该增强SSNR大于基准SSNR。
第三确定单元503,用于将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号。
图5所示的装置500可以确定输入的音频信号的特征,根据音频信号的特征,采用相应的方式确定增强SSNR,并采用该增强SSNR与VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低。
可选的,作为一个实施例,该第一确定单元501,具体用于根据该音频信号的子带SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在第一确定单元501根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,第一确定单元501,具体 用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,在第一确定单元501根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,第一确定单元501,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,在第一确定单元501根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,第一确定单元501,具体用于在该音频信号中的子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,第一确定单元501,具体用于在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。具体地,本领域技术人员可以理解,可以有多种用于检测音频信号是否为清音信号的方法。例如,可以通过检测该音频信号的时域过零率(Zero-Crossing Rate,ZCR)来确定该音频信号是否为清音信号。具体地,在该音频信号的ZCR大于ZCR阈值的情况下,确定该音频信号为清音信号,其中该ZCR阈值是通过大量实验确定的。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
该第三预设门限也是根据统计得到的。具体来说,从大量的噪声信号的子带SNR中确定第三预设门限,使得这些噪声信号中的绝大多数子带的子带SNR都小于该值。
第一数量、第二数量、第三数量和第四数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音样本中,统计高频端子带的子带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音样本中绝大多数的大于第一预设门限的高频端子带SNR的数量大于该第一数量。确 定第二数量的方法与确定第一数量的方法类似。第二数量可以与第一数量相同,也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音样本中,统计低频端子带的子带SNR大于第二预设门限的子带数量,从中确定第三数量,使得这些语音样本中绝大多数的大于第二预设门限的低频端子带SNR的数量大于该第三数量。对于第四数量,在大量的含有噪声的语音样本中,统计子带SNR大于第三预设门限的子带数量,从中确定第四数量,使得这些语音样本中绝大多数的大于第三预设门限的子带SNR的数量大于该第四数量。
进一步,第二确定单元502,具体用于确定该音频信号中各个子带的子带SNR的权重,其中该子带SNR大于第一预设门限的高频端子带的权重大于其他子带的子带SNR的权重,根据该音频信号中的各个子带的子带SNR的权重和各个子带的SNR,确定该增强SSNR。
可选的,作为一个实施例,第二确定单元502,具体用于确定该音频信号的基准SSNR,根据该音频信号的基准SSNR,确定增强SSNR。
该基准SSNR可以是采用公式1.1计算出来的SSNR。基准SSNR在计算时,计入SSNR的各个子带的子带SNR在SSNR中的权重相同。
可选的,作为另一个实施例,第二确定单元502,具体用于使用以下公式确定该增强SSNR:
SSNR′=x*SSNR+y,......................................................公式1.7
其中,SSNR表示该基准SSNR,SSNR′表示该增强SSNR,x和y表示增强参数。例如,x的取值可以为1.05,y的取值可以为1。本领域技术人员可以理解,x和y的取值还可以是其他合适的值,使得增强SSNR恰当的大于基准SSNR。
可选的,作为另一个实施例,第二确定单元502,具体用于使用以下公式确定该增强SSNR:
SSNR′=f(x)*SSNR+h(y),.............................................公式1.8
其中,SSNR表示该基准SSNR,SSNR′表示该增强SSNR,f(x)、h(y)表示增强函数。例如,f(x)和h(y)可以是与该音频信号的长时信噪比(Long-term SNR,LSNR)相关的函数,音频信号的长时信噪比为一段较长时间内的平均SNR或加权SNR。例如,当lsnr大于20时,f(lsnr)可以等于1.1,y(lsnr)可以等于2。当lsnr小于20且大于15时,f(lsnr)可以等于1.05, y(lsnr)可以等于1。当lsnr小于15时,f(lsnr)可以等于1,y(lsnr)可以等于0。本领域技术人员可以理解,f(x)和h(y)还可以是其他合适的形式,使得增强SSNR恰当的大于基准SSNR。
第三确定单元503,具体用于将该增强SSNR与语音活动检测VAD判决门限比较,根据比较结构确定该音频信号是否为活动信号。具体来说,如果该增强SSNR大于该VAD判决门限,则确定该音频信号为活动信号。如果该增强SSNR小于该VAD判决门限,则确定该音频信号为非活动信号。
可选的,作为另一个实施例,还可以使用预置算法减小基准VAD判决门限后获得的减小后的VAD判决门限,使用减小后的VAD判决门限确定该音频信号是否为活动信号。在此情况下,装置500还可以包括第四确定单元504。第四确定单元504用于使用预置算法减小该VAD判决门限,获得减小后的VAD判决门限。在此情况下,第三确定单元503,具体用于将该增强SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
图6是根据本发明实施例提供的另一装置的结构框图。图6所示的装置能够执行图3的各个步骤。如图6所示,装置600包括第一确定单元601、第二确定单元602和第三确定单元603。
第一确定单元601,用于确定输入的音频信号为待判断音频信号。
第二确定单元602,用于确定该音频信号中各个子带的子带信噪比SNR的权重,其中该子带SNR大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重,根据该音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定增强分段信噪比SSNR,其中该增强SSNR大于基准SSNR。
第三确定单元603,用于将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号。
图6所示的装置600可以确定输入的音频信号的特征,根据音频信号的特征,采用相应的方式确定增强SSNR,并采用该增强SSNR与VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低。
进一步,第一确定单元601,具体用于根据该音频信号的子带信噪比SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,第一确定单元601,具体用于在该音频信号 中子带信噪比SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,第一确定单元601,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
第一数量、第二数量和第三数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音清音样本帧中,统计高频端子带的子带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音清音样本帧中绝大多数的子带SNR大于第一预设门限的高频端子带的数量大于该第一数量。获取第二数量的方法与获取第一数量的方法类似。第二数量可以与第一数量相同,第二数量也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音清音样本帧中,统计低频端子带的子带SNR小于第二预设门限的子带数量,从中确定第三数量,使得这些语音清音样本帧中绝大多数的子带SNR小于第二预设门限的低频端子带的数量大于该第三数量。
图7是根据本发明实施例提供的装置的结构框图。图7所示的装置能够执行图1或图2的各个步骤。如图7所示,装置700包括处理器701和存储器702。其中,处理器701可以是通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现成可编程门阵列(Field Programmable Gate Array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本发明实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本发明实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成,或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存取存 储器(Random Access Memory,RAM)、闪存、只读存储器(Read-Only Memory,ROM)、可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器702,处理器701读取存储器702中的指令,结合其硬件完成上述方法的步骤。
处理器701,用于确定输入的音频信号为待判断音频信号。
处理器701,用于确定该音频信号的增强分段信噪比SSNR,其中该增强SSNR大于基准SSNR。
处理器701,用于将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号。
图7所示的装置700可以确定输入的音频信号的特征,根据音频信号的特征,采用相应的方式确定增强SSNR,并采用该增强SSNR与VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低。
可选的,作为一个实施例,该处理器701,具体用于根据该音频信号的子带SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在处理器701根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,处理器701,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,在处理器701根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,处理器701,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,在处理器701根据该音频信号的子带SNR确定该音频信号为待判断音频信号的情况下,处理器701,具体用于在该音频信号中的子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,处理器701,具体用于在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。具体地,本领域技术人员可以理解,可以有多种用于检测音频信号是否为清音信号的方法。例如,可以通过检测该音频信号的时域过零率(Zero-Crossing Rate,ZCR) 来确定该音频信号是否为清音信号。具体地,在该音频信号的ZCR大于ZCR阈值的情况下,确定该音频信号为清音信号,其中该ZCR阈值是通过大量实验确定的。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
该第三预设门限也是根据统计得到的。具体来说,从大量的噪声信号的子带SNR中确定第三预设门限,使得这些噪声信号中的绝大多数子带的子带SNR都小于该值。
第一数量、第二数量、第三数量和第四数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音样本中,统计高频端子带的子带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音样本中绝大多数的大于第一预设门限的高频端子带SNR的数量大于该第一数量。确定第二数量的方法与确定第一数量的方法类似。第二数量可以与第一数量相同,也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音样本中,统计低频端子带的子带SNR大于第二预设门限的子带数量,从中确定第三数量,使得这些语音样本中绝大多数的大于第二预设门限的低频端子带SNR的数量大于该第三数量。对于第四数量,在大量的含有噪声的语音样本中,统计子带SNR大于第三预设门限的子带数量,从中确定第四数量,使得这些语音样本中绝大多数的大于第三预设门限的子带SNR的数量大于该第四数量。
进一步,处理器701,具体用于确定该音频信号中各个子带的子带SNR的权重,其中该子带SNR大于第一预设门限的高频端子带的权重大于其他子带的子带SNR的权重,根据该音频信号中的各个子带的子带SNR的权重和各个子带的SNR,确定该增强SSNR。
可选的,作为一个实施例,处理器701,具体用于确定该音频信号的基准SSNR,根据该音频信号的基准SSNR,确定增强SSNR。
该基准SSNR可以是采用公式1.1计算出来的SSNR。基准SSNR在计 算时,计入SSNR的各个子带的子带SNR在SSNR中的权重相同。
可选的,作为另一个实施例,处理器701,具体用于使用以下公式确定该增强SSNR:
SSNR′=x*SSNR+y,......................................................公式1.7
其中,SSNR表示该基准SSNR,SSNR′表示该增强SSNR,x和y表示增强参数。例如,x的取值可以为1.07,y的取值可以为1。本领域技术人员可以理解,x和y的取值还可以是其他合适的值,使得增强SSNR恰当的大于基准SSNR。
可选的,作为另一个实施例,处理器701,具体用于使用以下公式确定该增强SSNR:
SSNR′=f(x)*SSNR+h(y),.............................................公式1.8
其中,SSNR表示该基准SSNR,SSNR′表示该增强SSNR,f(x)、h(y)表示增强函数。例如,f(x)和h(y)可以是与该音频信号的长时信噪比(Long-term SNR,LSNR)相关的函数,音频信号的长时信噪比为一段较长时间内的平均SNR或加权SNR。例如,当lsnr大于20时,f(lsnr)可以等于1.1,y(lsnr)可以等于2。当lsnr小于20且大于17时,f(lsnr)可以等于1.07,y(lsnr)可以等于1。当lsnr小于17时,f(lsnr)可以等于1,y(lsnr)可以等于0。本领域技术人员可以理解,f(x)和h(y)还可以是其他合适的形式,使得增强SSNR恰当的大于基准SSNR。
处理器701,具体用于将该增强SSNR与语音活动检测VAD判决门限比较,根据比较结构确定该音频信号是否为活动信号。具体来说,如果该增强SSNR大于该VAD判决门限,则确定该音频信号为活动信号。如果该增强SSNR小于该VAD判决门限,则确定该音频信号为非活动信号。
可选的,作为另一个实施例,还可以使用预置算法减小基准VAD判决门限后获得的减小后的VAD判决门限,使用减小后的VAD判决门限确定该音频信号是否为活动信号。在此情况下,处理器701还可以用于使用预置算法减小该VAD判决门限,获得减小后的VAD判决门限。在此情况下,处理器701具体用于将该增强SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
图8是根据本发明实施例提供的另一装置的结构框图。图8所示的装置能够执行图3的各个步骤。如图8所示,装置800包括处理器801和存储器 802。其中,处理器801可以是通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现成可编程门阵列(Field Programmable Gate Array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本发明实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本发明实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成,或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存取存储器(Random Access Memory,RAM)、闪存、只读存储器(Read-Only Memory,ROM)、可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器802,处理器801读取存储器802中的指令,结合其硬件完成上述方法的步骤。
处理器801,用于确定输入的音频信号为待判断音频信号。
处理器801,用于确定该音频信号中各个子带的子带信噪比SNR的权重,其中该子带SNR大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重,根据该音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定增强分段信噪比SSNR,其中该增强SSNR大于基准SSNR。
处理器801,用于将该增强SSNR与语音活动检测VAD判决门限比较,确定该音频信号是否为活动信号。
图8所示的装置800可以确定输入的音频信号的特征,根据音频信号的特征,采用相应的方式确定增强SSNR,并采用该增强SSNR与VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低。
进一步,处理器801,具体用于根据该音频信号的子带信噪比SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,处理器801,具体用于在该音频信号中子带信噪比SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为另一个实施例,处理器801,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下, 确定该音频信号为待判断音频信号。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
第一数量、第二数量和第三数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音清音样本帧中,统计高频端子带的子带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音清音样本帧中绝大多数的子带SNR大于第一预设门限的高频端子带的数量大于该第一数量。获取第二数量的方法与获取第一数量的方法类似。第二数量可以与第一数量相同,第二数量也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音清音样本帧中,统计低频端子带的子带SNR小于第二预设门限的子带数量,从中确定第三数量,使得这些语音清音样本帧中绝大多数的子带SNR小于第二预设门限的低频端子带的数量大于该第三数量。
图9是根据本发明实施例提供的另一装置的结构框图。图9所示的装置900可以执行图4的各个步骤。如图9所示,装置900包括:第一确定单元901、第二确定单元902、第三确定单元903和第四确定单元904。
第一确定单元901,用于确定输入的音频信号为待判断音频信号。
第二确定单元902,用于获取该音频信号的基准SSNR。
具体地,该基准SSNR可以是采用公式1.1计算出来的SSNR。
第三确定单元903,用于使用预置算法减小基准VAD判决门限,获得减小后的VAD判决门限。
具体地,基准VAD判决门限可以是默认的VAD判决门限,该基准VAD判决门限可以是预先存储的,也可以是临时计算获得,其中基准VAD判决门限的计算可以采用现有公知技术。在使用预置算法减小基准VAD判决门限时,该预置算法可以是将基准VAD判决门限乘于一个小于1的系数,也可以采用其他算法,本发明实施例并不限定所采用的具体算法。该预置算法可以适当减小VAD判决门限,使得增强SSNR大于该减小后的VAD判决门限,从而可以使得活动信号被漏减的比例降低。
第四确定单元904,用于将该基准SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
可选的,作为一个实施例,第一确定单元901,具体用于根据该音频信号的SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在第一确定单元901根据该音频信号的SNR确定该音频信号为待判断音频信号的情况下,第一确定单元901,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在第一确定单元901根据该音频信号的SNR确定该音频信号为待判断音频信号的情况下,第一确定单元901,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在第一确定单元901根据该音频信号的SNR确定该音频信号为待判断音频信号的情况下,第一确定单元901,具体用于在该音频信号中中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,第一确定单元901,具体用于在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。具体地,本领域技术人员可以理解,可以有多种用于检测音频信号是否为清音信号的方法。例如,可以通过检测该音频信号的时域过零率(Zero-Crossing Rate,ZCR)来确定该音频信号是否为清音信号。具体地,在该音频信号的ZCR大于ZCR阈值的情况下,确定该音频信号为清音信号,其中该ZCR阈值是通过大量实验确定的。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
该第三预设门限也是根据统计得到的。具体来说,从大量的噪声信号的 子带SNR中确定第三预设门限,使得这些噪声信号中的绝大多数子带的子带SNR都小于该值。
第一数量、第二数量、第三数量和第四数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音样本中,统计高频端子带的子带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音样本中绝大多数的大于第一预设门限的高频端子带SNR的数量大于该第一数量。确定第二数量的方法与确定第一数量的方法类似。第二数量可以与第一数量相同,也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音样本中,统计低频端子带的子带SNR大于第二预设门限的子带数量,从中确定第三数量,使得这些语音样本中绝大多数的大于第二预设门限的低频端子带SNR的数量大于该第三数量。对于第四数量,在大量的含有噪声的语音样本中,统计子带SNR大于第三预设门限的子带数量,从中确定第四数量,使得这些语音样本中绝大多数的大于第三预设门限的子带SNR的数量大于该第四数量。
图9所示的装置900可以确定输入的音频信号的特征,根据音频信号的特征,减小基准VAD判决门限,并采用SSNR与减小后的VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低。
图10是根据本发明实施例提供的另一装置的结构框图。图10所示的装置1000可以执行图4的各个步骤。如图10所示,装置1000包括:处理器1001和存储器1002。其中,处理器1001可以是通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现成可编程门阵列(Field Programmable Gate Array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本发明实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本发明实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成,或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存取存储器(Random Access Memory,RAM)、闪存、只读存储器(Read-Only Memory,ROM)、可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器1002,处理器1001读取存储器1002中的指令,结合其硬件完成上述方法的步骤。
处理器1001,用于确定输入的音频信号为待判断音频信号。
处理器1001,用于获取该音频信号的基准SSNR。
具体地,该基准SSNR可以是采用公式1.1计算出来的SSNR。
处理器1001,用于使用预置算法减小基准VAD判决门限,获得减小后的VAD判决门限。
具体地,基准VAD判决门限可以是默认的VAD判决门限,该基准VAD判决门限可以是预先存储的,也可以是临时计算获得,其中基准VAD判决门限的计算可以采用现有公知技术。在使用预置算法减小基准VAD判决门限时,该预置算法可以是将基准VAD判决门限乘于一个小于1的系数,也可以采用其他算法,本发明实施例并不限定所采用的具体算法。该预置算法可以适当减小VAD判决门限,使得增强SSNR大于该减小后的VAD判决门限,从而可以使得活动信号被漏减的比例降低。
处理器1001,用于将该基准SSNR与该减小后的VAD判决门限进行比较,确定该音频信号是否为活动信号。
可选的,作为一个实施例,处理器1001,具体用于根据该音频信号的SNR,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在处理器1001根据该音频信号的SNR确定该音频信号为待判断音频信号的情况下,处理器1001,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在处理器1001根据该音频信号的SNR确定该音频信号为待判断音频信号的情况下,处理器1001,具体用于在该音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且该音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,在处理器1001根据该音频信号的SNR确定该音频信号为待判断音频信号的情况下,处理器1001,具体用于在该音频信号中中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定该音频信号为待判断音频信号。
可选的,作为一个实施例,处理器1001,具体用于在确定该音频信号为清音信号的情况下,确定该音频信号为待判断音频信号。具体地,本领域技 术人员可以理解,可以有多种用于检测音频信号是否为清音信号的方法。例如,可以通过检测该音频信号的时域过零率(Zero-Crossing Rate,ZCR)来确定该音频信号是否为清音信号。具体地,在该音频信号的ZCR大于ZCR阈值的情况下,确定该音频信号为清音信号,其中该ZCR阈值是通过大量实验确定的。
该第一预设门限和该第二预设门限可以是根据大量的语音样本统计得到的。具体来说,在大量含有背景噪声的语音清音样本中,统计高频端子带的子带SNR,从中确定第一预设门限,使得这些清音样本中绝大多数的高频端子带的子带SNR均大于该门限。类似的,在这些语音清音样本中统计低频端子带的子带SNR,从中确定第二预设门限,使得这些语音清音样本中的绝大多数低频端子带的子带SNR均小于该门限。
该第三预设门限也是根据统计得到的。具体来说,从大量的噪声信号的子带SNR中确定第三预设门限,使得这些噪声信号中的绝大多数子带的子带SNR都小于该值。
第一数量、第二数量、第三数量和第四数量也是根据统计得到的。以第一数量为例,在大量的含有噪声的语音样本中,统计高频端子带的子带SNR大于第一预设门限的子带数量,从中确定第一数量,使得这些语音样本中绝大多数的大于第一预设门限的高频端子带SNR的数量大于该第一数量。确定第二数量的方法与确定第一数量的方法类似。第二数量可以与第一数量相同,也可以与第一数量不同。类似的,对于第三数量,在大量的含有噪声的语音样本中,统计低频端子带的子带SNR大于第二预设门限的子带数量,从中确定第三数量,使得这些语音样本中绝大多数的大于第二预设门限的低频端子带SNR的数量大于该第三数量。对于第四数量,在大量的含有噪声的语音样本中,统计子带SNR大于第三预设门限的子带数量,从中确定第四数量,使得这些语音样本中绝大多数的大于第三预设门限的子带SNR的数量大于该第四数量。
图10所示的装置1000可以确定输入的音频信号的特征,根据音频信号的特征,减小基准VAD判决门限,并采用SSNR与减小后的VAD判决门限进行比较,这样可以使得活动信号被漏检比例降低。
本领域普通技术人员可以意识到,结合本文中所公开的实施例描述的各示例的单元及算法步骤,能够以电子硬件、或者计算机软件和电子硬件的结 合来实现。这些功能究竟以硬件还是软件方式来执行,取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能,但是这种实现不应认为超出本发明的范围。
所属领域的技术人员可以清楚地了解到,为描述的方便和简洁,上述描述的系统、装置和单元的具体工作过程,可以参考前述方法实施例中的对应过程,在此不再赘述。
在本申请所提供的几个实施例中,应该理解到,所揭露的系统、装置和方法,可以通过其它的方式实现。例如,以上所描述的装置实施例仅仅是示意性的,例如,所述单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,例如多个单元或组件可以结合或者可以集成到另一个系统,或一些特征可以忽略,或不执行。另一点,所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口,装置或单元的间接耦合或通信连接,可以是电性,机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外,在本发明各个实施例中的各功能单元可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。基于这样的理解,本发明的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)或处理器(processor)执行本发明各个实施例所述方法的全部或部分步骤。而前述的存储介质包括:U盘、移动硬盘、只读存储器(ROM,Read-Only Memory)、随机存取存储器(RAM,Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述,仅为本发明的具体实施方式,但本发明的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本发明揭露的技术范围内,可轻易 想到的变化或替换,都应涵盖在本发明的保护范围之内,因此本发明的保护范围应以权利要求的保护范围为准。

Claims (42)

  1. 一种检测音频信号的方法,其特征在于,所述方法包括:
    确定输入的音频信号为待判断音频信号;
    确定所述音频信号的增强分段信噪比SSNR,其中所述增强SSNR大于基准SSNR;
    将所述增强SSNR与语音活动检测VAD判决门限进行比较,确定所述音频信号是否为活动信号。
  2. 如权利要求1所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    根据所述音频信号的子带信噪比SNR,确定所述音频信号为待判断音频信号。
  3. 如权利要求2所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定所述音频信号为待判断音频信号。
  4. 如权利要求2所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且所述音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定所述音频信号为待判断音频信号。
  5. 如权利要求2所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在所述音频信号中中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定所述音频信号为待判断音频信号。
  6. 如权利要求1所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在确定所述音频信号为清音信号的情况下,确定所述音频信号为待判断音频信号。
  7. 如权利要求3或4所述的方法,其特征在于,所述确定所述音频信号的增强分段信噪比SSNR,包括:
    确定所述音频信号中各个子带的子带SNR的权重,其中所述子带SNR 大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重;
    根据所述音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定所述增强SSNR。
  8. 如权利要求1-6中任一项所述的方法,其特征在于,所述确定所述音频信号的增强分段信噪比SSNR,包括:
    确定所述音频信号的基准SSNR;
    根据所述音频信号的基准SSNR,确定增强SSNR。
  9. 如权利要求8所述的方法,其特征在于,所述根据所述音频信号的基准SSNR,确定增强SSNR,包括:
    使用以下公式确定所述增强SSNR:
    SSNR′=x*SSNR+y,
    其中,SSNR表示所述基准SSNR,SSNR′表示所述增强SSNR,x和y表示增强参数。
  10. 如权利要求8所述的方法,其特征在于,所述根据所述音频信号的基准SSNR,确定增强SSNR,包括:
    使用以下公式确定所述增强SSNR:
    SSNR′=f(x)*SSNR+h(y),
    其中,SSNR表示所述基准SSNR,SSNR′表示所述增强SSNR,f(x)、h(y)表示增强函数。
  11. 如权利要求1至10任一所述的方法,其特征在于,所述将所述增强SSNR与语音活动检测VAD判决门限进行比较前进一步包括:
    使用预置算法减小所述VAD判决门限,获得减小后的VAD判决门限;
    所述将所述增强SSNR与语音活动检测VAD判决门限比较,确定所述音频信号是否为活动信号具体包括:
    将所述增强SSNR与所述减小后的VAD判决门限进行比较,确定所述音频信号是否为活动信号。
  12. 一种检测音频信号的方法,其特征在于,所述方法包括:
    确定输入的音频信号为待判断音频信号;
    确定所述音频信号中各个子带的子带信噪比SNR的权重,其中所述子带SNR大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的 子带SNR的权重;
    根据所述音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定增强分段信噪比SSNR,其中所述增强SSNR大于基准SSNR;
    将所述增强SSNR与语音活动检测VAD判决门限比较,确定所述音频信号是否为活动信号。
  13. 如权利要求12所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    根据所述音频信号的子带SNR,确定所述音频信号为待判断音频信号。
  14. 如权利要求13所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定所述音频信号为待判断音频信号。
  15. 如权利要求13所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且所述音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定所述音频信号为待判断音频信号。
  16. 一种检测音频信号的方法,其特征在于,所述方法包括:
    确定输入的音频信号为待判断音频信号;
    获取所述音频信号的基准分段信噪比SSNR;
    使用预置算法减小基准语音活动检测VAD判决门限,获得减小后的VAD判决门限;
    将所述基准SSNR与所述减小后的VAD判决门限进行比较,确定所述音频信号是否为活动信号。
  17. 如权利要求16所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    根据所述音频信号的子带信噪比SNR,确定所述音频信号为待判断音频信号。
  18. 如权利要求17所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大 于第一数量的情况下,确定所述音频信号为待判断音频信号。
  19. 如权利要求17所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且所述音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定所述音频信号为待判断音频信号。
  20. 如权利要求17所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在所述音频信号中中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定所述音频信号为待判断音频信号。
  21. 如权利要求16所述的方法,其特征在于,所述确定输入的音频信号为待判断音频信号,包括:
    在确定所述音频信号为清音信号的情况下,确定所述音频信号为待判断音频信号。
  22. 一种装置,其特征在于,所述装置包括:
    第一确定单元,用于确定输入的音频信号为待判断音频信号;
    第二确定单元,用于确定所述音频信号的增强分段信噪比SSNR,其中所述增强SSNR大于基准SSNR;
    第三确定单元,用于将所述增强SSNR与语音活动检测VAD判决门限比较,确定所述音频信号是否为活动信号。
  23. 如权利要求22所述的装置,其特征在于,所述第一确定单元,具体用于根据所述音频信号的子带信噪比SNR,确定所述音频信号为待判断音频信号。
  24. 如权利要求23所述的装置,其特征在于,所述第一确定单元,具体用于在所述音频信号中子带信噪比SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定所述音频信号为待判断音频信号。
  25. 如权利要求23所述的装置,其特征在于,所述第一确定单元,具体用于在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且所述音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定所述音频信号为待判断音频信号。
  26. 如权利要求23所述的装置,其特征在于,所述第一确定单元,具 体用于在所述音频信号中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定所述音频信号为待判断音频信号。
  27. 如权利要求22所述的装置,其特征在于,所述第一确定单元,具体用于在确定所述音频信号为清音信号的情况下,确定所述音频信号为待判断音频信号。
  28. 如权利要求24或25所述的装置,其特征在于,所述第二确定单元,具体用于确定所述音频信号中各个子带的子带SNR的权重,其中所述子带SNR大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重,根据所述音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定所述增强SSNR。
  29. 如权利要求22-27中任一项所述的装置,其特征在于,第二确定单元,具体用于确定所述音频信号的基准SSNR,根据所述音频信号的基准SSNR,确定所述增强SSNR。
  30. 如权利要求29所述的装置,其特征在于,所述第二确定单元,具体用于使用以下公式确定所述增强SSNR:
    SSNR′=x*SSNR+y,
    其中,SSNR表示所述基准SSNR,SSNR′表示所述增强SSNR,x和y表示增强参数。
  31. 如权利要求29所述的装置,其特征在于,所述第二确定单元,具体用于使用以下公式确定所述增强SSNR:
    SSNR′=f(x)*SSNR+h(y),
    其中,SSNR表示所述基准SSNR,SSNR′表示所述增强SSNR,f(x)、h(y)表示增强函数。
  32. 如权利要求22至31中任一项所述的装置,其特征在于,所述装置还包括第四确定单元;
    所述第四确定单元,用于使用预置算法减小所述VAD判决门限,获得减小后的VAD判决门限;
    所述第三确定单元,具体用于将所述增强SSNR与所述减小后的VAD判决门限进行比较,确定所述音频信号是否为活动信号。
  33. 一种装置,其特征在于,所述装置包括:
    第一确定单元,用于确定输入的音频信号为待判断音频信号;
    第二确定单元,用于确定所述音频信号中各个子带的子带信噪比SNR的权重,其中所述子带SNR大于第一预设门限的高频端子带的子带SNR的权重大于其他子带的子带SNR的权重,根据所述音频信号中的各个子带的子带SNR的权重和各个子带的子带SNR,确定增强分段信噪比SSNR,其中所述增强SSNR大于基准SSNR;
    第三确定单元,用于将所述增强SSNR与语音活动检测VAD判决门限比较,确定所述音频信号是否为活动信号。
  34. 如权利要求33所述的装置,其特征在于,所述第一确定单元,具体用于根据所述音频信号的子带信噪比SNR,确定所述音频信号为待判断音频信号。
  35. 如权利要求34所述的装置,其特征在于,所述第一确定单元,具体用于在所述音频信号中子带信噪比SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定所述音频信号为待判断音频信号。
  36. 如权利要求34所述的装置,其特征在于,所述第一确定单元,具体用于在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且所述音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定所述音频信号为待判断音频信号。
  37. 一种装置,其特征在于,所述装置包括:
    第一确定单元,用于确定输入的音频信号为待判断音频信号;
    第二确定单元,用于获取所述音频信号的基准分段信噪比SSNR;
    第三确定单元,用于使用预置算法减小基准语音活动检测VAD判决门限,获得减小后的VAD判决门限;
    第四确定单元,用于将所述基准SSNR与所述减小后的VAD判决门限进行比较,确定所述音频信号是否为活动信号。
  38. 如权利要求37所述的装置,其特征在于,所述第一确定单元,具体用于根据所述音频信号的子带信噪比SNR,确定所述音频信号为待判断音频信号。
  39. 如权利要求38所述的装置,其特征在于,所述第一确定单元,具体用于在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第一数量的情况下,确定所述音频信号为待判断音频信号。
  40. 如权利要求38所述的装置,其特征在于,所述第一确定单元,具 体用于在所述音频信号中子带SNR大于第一预设门限的高频端子带的数量大于第二数量且所述音频信号中子带SNR小于第二预设门限的低频端子带的数量大于第三数量情况下,确定所述音频信号为待判断音频信号。
  41. 如权利要求38所述的装置,其特征在于,所述第一确定单元,具体用于在所述音频信号中中子带SNR的值大于第三预设门限的子带的数量大于第四数量的情况下,确定所述音频信号为待判断音频信号。
  42. 如权利要求37所述的装置,其特征在于,所述第一确定单元,具体用于在确定所述音频信号为清音信号的情况下,确定所述音频信号为待判断音频信号。
PCT/CN2014/092694 2014-03-12 2014-12-01 检测音频信号的方法和装置 Ceased WO2015135344A1 (zh)

Priority Applications (15)

Application Number Priority Date Filing Date Title
MYPI2016703030A MY193521A (en) 2014-03-12 2014-12-01 Method for detecting audio signal and apparatus
EP14885786.5A EP3118852B1 (en) 2014-03-12 2014-12-01 Method and device for detecting audio signal
MX2016011750A MX355828B (es) 2014-03-12 2014-12-01 Método y aparato para la detección de señales de audio.
KR1020187021506A KR102005009B1 (ko) 2014-03-12 2014-12-01 오디오 신호를 검출하는 방법 및 장치
EP19197660.4A EP3660845B1 (en) 2014-03-12 2014-12-01 Method for detecting audio signal and apparatus
AU2014386442A AU2014386442B9 (en) 2014-03-12 2014-12-01 Method for detecting audio signal and apparatus
SG11201607052SA SG11201607052SA (en) 2014-03-12 2014-12-01 Method for detecting audio signal and apparatus
CA2940487A CA2940487C (en) 2014-03-12 2014-12-01 Method for detecting audio signal and apparatus
RU2016139717A RU2666337C2 (ru) 2014-03-12 2014-12-01 Способ обнаружения звукового сигнала и устройство
JP2016556770A JP6493889B2 (ja) 2014-03-12 2014-12-01 音声信号を検出するための方法および装置
KR1020167025280A KR101884220B1 (ko) 2014-03-12 2014-12-01 오디오 신호를 검출하는 방법 및 장치
ES14885786T ES2787894T3 (es) 2014-03-12 2014-12-01 Método y dispositivo para detectar la señal de audio
US15/262,263 US10304478B2 (en) 2014-03-12 2016-09-12 Method for detecting audio signal and apparatus
US16/391,893 US10818313B2 (en) 2014-03-12 2019-04-23 Method for detecting audio signal and apparatus
US16/901,846 US11417353B2 (en) 2014-03-12 2020-06-15 Method for detecting audio signal and apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410090386.XA CN104916292B (zh) 2014-03-12 2014-03-12 检测音频信号的方法和装置
CN201410090386.X 2014-03-12

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US15/262,263 Continuation US10304478B2 (en) 2014-03-12 2016-09-12 Method for detecting audio signal and apparatus

Publications (1)

Publication Number Publication Date
WO2015135344A1 true WO2015135344A1 (zh) 2015-09-17

Family

ID=54070889

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/092694 Ceased WO2015135344A1 (zh) 2014-03-12 2014-12-01 检测音频信号的方法和装置

Country Status (14)

Country Link
US (3) US10304478B2 (zh)
EP (2) EP3118852B1 (zh)
JP (2) JP6493889B2 (zh)
KR (2) KR102005009B1 (zh)
CN (3) CN107086043B (zh)
AU (1) AU2014386442B9 (zh)
CA (1) CA2940487C (zh)
ES (2) ES2926360T3 (zh)
MX (1) MX355828B (zh)
MY (1) MY193521A (zh)
PT (2) PT3660845T (zh)
RU (1) RU2666337C2 (zh)
SG (1) SG11201607052SA (zh)
WO (1) WO2015135344A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107040359A (zh) * 2017-05-08 2017-08-11 海能达通信股份有限公司 一种语音呼叫过程中携带随路信令的方法、装置及设备

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107086043B (zh) * 2014-03-12 2020-09-08 华为技术有限公司 检测音频信号的方法和装置
CN113038353B (zh) * 2016-04-29 2022-08-09 荣耀终端有限公司 一种语音输入异常的确定方法、装置、终端以及存储介质
CN107393558B (zh) * 2017-07-14 2020-09-11 深圳永顺智信息科技有限公司 语音活动检测方法及装置
CN107393553B (zh) * 2017-07-14 2020-12-22 深圳永顺智信息科技有限公司 用于语音活动检测的听觉特征提取方法
CN107393559B (zh) * 2017-07-14 2021-05-18 深圳永顺智信息科技有限公司 检校语音检测结果的方法及装置
CN107393550B (zh) * 2017-07-14 2021-03-19 深圳永顺智信息科技有限公司 语音处理方法及装置
US11783809B2 (en) * 2020-10-08 2023-10-10 Qualcomm Incorporated User voice activity detection using dynamic classifier
EP4404196A4 (en) 2021-11-09 2025-01-22 Samsung Electronics Co., Ltd. ELECTRONIC DEVICE FOR CONTROLLING BEAM FORMING AND OPERATING METHOD THEREFOR
US20240304203A1 (en) * 2023-03-06 2024-09-12 Nvidia Corporation Noise reduction using voice activity detection in audio processing systems and applications

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6453291B1 (en) * 1999-02-04 2002-09-17 Motorola, Inc. Apparatus and method for voice activity detection in a communication system
EP2113908A1 (en) * 2008-04-30 2009-11-04 QNX Software Systems (Wavemakers), Inc. Robust downlink speech and noise detector
CN102044243A (zh) * 2009-10-15 2011-05-04 华为技术有限公司 语音激活检测方法与装置、编码器
CN102044242A (zh) * 2009-10-15 2011-05-04 华为技术有限公司 语音激活检测方法、装置和电子设备
CN102576528A (zh) * 2009-10-19 2012-07-11 瑞典爱立信有限公司 用于语音活动检测的检测器和方法
CN102959625A (zh) * 2010-12-24 2013-03-06 华为技术有限公司 自适应地检测输入音频信号中的话音活动的方法和设备

Family Cites Families (48)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS59182498A (ja) * 1983-04-01 1984-10-17 日本電気株式会社 音声検出回路
JPS63259596A (ja) * 1987-04-16 1988-10-26 株式会社日立製作所 音声区間検出方式
CA2153170C (en) * 1993-11-30 2000-12-19 At&T Corp. Transmitted noise reduction in communications systems
FI100840B (fi) * 1995-12-12 1998-02-27 Nokia Mobile Phones Ltd Kohinanvaimennin ja menetelmä taustakohinan vaimentamiseksi kohinaises ta puheesta sekä matkaviestin
US5991718A (en) * 1998-02-27 1999-11-23 At&T Corp. System and method for noise threshold adaptation for voice activity detection in nonstationary noise environments
US6466906B2 (en) * 1999-01-06 2002-10-15 Dspc Technologies Ltd. Noise padding and normalization in dynamic time warping
US6324509B1 (en) * 1999-02-08 2001-11-27 Qualcomm Incorporated Method and apparatus for accurate endpointing of speech in the presence of noise
JP2001236085A (ja) * 2000-02-25 2001-08-31 Matsushita Electric Ind Co Ltd 音声区間検出装置、定常雑音区間検出装置、非定常雑音区間検出装置、及び雑音区間検出装置
JP3588030B2 (ja) * 2000-03-16 2004-11-10 三菱電機株式会社 音声区間判定装置及び音声区間判定方法
US6898566B1 (en) * 2000-08-16 2005-05-24 Mindspeed Technologies, Inc. Using signal to noise ratio of a speech signal to adjust thresholds for extracting speech parameters for coding the speech signal
CN1175398C (zh) * 2000-11-18 2004-11-10 中兴通讯股份有限公司 一种从噪声环境中识别出语音和音乐的声音活动检测方法
WO2002080148A1 (en) * 2001-03-28 2002-10-10 Mitsubishi Denki Kabushiki Kaisha Noise suppressor
US7941313B2 (en) * 2001-05-17 2011-05-10 Qualcomm Incorporated System and method for transmitting speech activity information ahead of speech features in a distributed voice recognition system
US7203643B2 (en) 2001-06-14 2007-04-10 Qualcomm Incorporated Method and apparatus for transmitting speech activity in distributed voice recognition systems
US6937980B2 (en) * 2001-10-02 2005-08-30 Telefonaktiebolaget Lm Ericsson (Publ) Speech recognition using microphone antenna array
JP4281349B2 (ja) * 2001-12-25 2009-06-17 パナソニック株式会社 電話装置
US7024353B2 (en) * 2002-08-09 2006-04-04 Motorola, Inc. Distributed speech recognition with back-end voice activity detection apparatus and method
US7146315B2 (en) * 2002-08-30 2006-12-05 Siemens Corporate Research, Inc. Multichannel voice detection in adverse environments
US7162420B2 (en) * 2002-12-10 2007-01-09 Liberato Technologies, Llc System and method for noise reduction having first and second adaptive filters
JP4490090B2 (ja) * 2003-12-25 2010-06-23 株式会社エヌ・ティ・ティ・ドコモ 有音無音判定装置および有音無音判定方法
CA2454296A1 (en) 2003-12-29 2005-06-29 Nokia Corporation Method and device for speech enhancement in the presence of background noise
US8340309B2 (en) * 2004-08-06 2012-12-25 Aliphcom, Inc. Noise suppressing multi-microphone headset
CN100369113C (zh) * 2004-12-31 2008-02-13 中国科学院自动化研究所 利用增益自适应提高语音识别率的方法
US8175877B2 (en) * 2005-02-02 2012-05-08 At&T Intellectual Property Ii, L.P. Method and apparatus for predicting word accuracy in automatic speech recognition systems
WO2007091956A2 (en) 2006-02-10 2007-08-16 Telefonaktiebolaget Lm Ericsson (Publ) A voice detector and a method for suppressing sub-bands in a voice detector
US8032370B2 (en) * 2006-05-09 2011-10-04 Nokia Corporation Method, apparatus, system and software product for adaptation of voice activity detection parameters based on the quality of the coding modes
US8311814B2 (en) * 2006-09-19 2012-11-13 Avaya Inc. Efficient voice activity detector to detect fixed power signals
CN101197130B (zh) * 2006-12-07 2011-05-18 华为技术有限公司 声音活动检测方法和声音活动检测器
US7769585B2 (en) * 2007-04-05 2010-08-03 Avidyne Corporation System and method of voice activity detection in noisy environments
CN101320559B (zh) * 2007-06-07 2011-05-18 华为技术有限公司 一种声音激活检测装置及方法
US8954324B2 (en) * 2007-09-28 2015-02-10 Qualcomm Incorporated Multiple microphone voice activity detector
KR101335417B1 (ko) 2008-03-31 2013-12-05 (주)트란소노 노이지 음성 신호의 처리 방법과 이를 위한 장치 및 컴퓨터판독 가능한 기록매체
US8768690B2 (en) * 2008-06-20 2014-07-01 Qualcomm Incorporated Coding scheme selection for low-bit-rate applications
WO2010091339A1 (en) 2009-02-06 2010-08-12 University Of Ottawa Method and system for noise reduction for speech enhancement in hearing aid
JP5337530B2 (ja) * 2009-02-25 2013-11-06 京セラ株式会社 無線基地局および無線通信方法
KR20110001130A (ko) * 2009-06-29 2011-01-06 삼성전자주식회사 가중 선형 예측 변환을 이용한 오디오 신호 부호화 및 복호화 장치 및 그 방법
WO2011049515A1 (en) * 2009-10-19 2011-04-28 Telefonaktiebolaget Lm Ericsson (Publ) Method and voice activity detector for a speech encoder
US8898058B2 (en) * 2010-10-25 2014-11-25 Qualcomm Incorporated Systems, methods, and apparatus for voice activity detection
EP3252771B1 (en) 2010-12-24 2019-05-01 Huawei Technologies Co., Ltd. A method and an apparatus for performing a voice activity detection
WO2012083552A1 (en) * 2010-12-24 2012-06-28 Huawei Technologies Co., Ltd. Method and apparatus for voice activity detection
US9099098B2 (en) * 2012-01-20 2015-08-04 Qualcomm Incorporated Voice activity detection in presence of background noise
WO2013118192A1 (ja) * 2012-02-10 2013-08-15 三菱電機株式会社 雑音抑圧装置
JP5862349B2 (ja) * 2012-02-16 2016-02-16 株式会社Jvcケンウッド ノイズ低減装置、音声入力装置、無線通信装置、およびノイズ低減方法
CN103325380B (zh) * 2012-03-23 2017-09-12 杜比实验室特许公司 用于信号增强的增益后处理
US20130282373A1 (en) 2012-04-23 2013-10-24 Qualcomm Incorporated Systems and methods for audio signal processing
US9524735B2 (en) * 2014-01-31 2016-12-20 Apple Inc. Threshold adaptation in two-channel noise estimation and voice activity detection
CN107086043B (zh) * 2014-03-12 2020-09-08 华为技术有限公司 检测音频信号的方法和装置
US9775113B2 (en) * 2014-12-11 2017-09-26 Mediatek Inc. Voice wakeup detecting device with digital microphone and associated method

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6453291B1 (en) * 1999-02-04 2002-09-17 Motorola, Inc. Apparatus and method for voice activity detection in a communication system
EP2113908A1 (en) * 2008-04-30 2009-11-04 QNX Software Systems (Wavemakers), Inc. Robust downlink speech and noise detector
CN102044243A (zh) * 2009-10-15 2011-05-04 华为技术有限公司 语音激活检测方法与装置、编码器
CN102044242A (zh) * 2009-10-15 2011-05-04 华为技术有限公司 语音激活检测方法、装置和电子设备
CN102576528A (zh) * 2009-10-19 2012-07-11 瑞典爱立信有限公司 用于语音活动检测的检测器和方法
CN102959625A (zh) * 2010-12-24 2013-03-06 华为技术有限公司 自适应地检测输入音频信号中的话音活动的方法和设备

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3118852A4 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107040359A (zh) * 2017-05-08 2017-08-11 海能达通信股份有限公司 一种语音呼叫过程中携带随路信令的方法、装置及设备
CN107040359B (zh) * 2017-05-08 2021-01-19 海能达通信股份有限公司 一种语音呼叫过程中携带随路信令的方法、装置及设备

Also Published As

Publication number Publication date
US20160379670A1 (en) 2016-12-29
JP6793706B2 (ja) 2020-12-02
JP2019053321A (ja) 2019-04-04
CN104916292B (zh) 2017-05-24
US10304478B2 (en) 2019-05-28
US10818313B2 (en) 2020-10-27
PT3118852T (pt) 2020-03-06
AU2014386442A1 (en) 2016-09-08
MY193521A (en) 2022-10-17
CN107293287B (zh) 2021-10-26
AU2014386442B9 (en) 2017-11-23
SG11201607052SA (en) 2016-10-28
AU2014386442B2 (en) 2017-11-02
RU2666337C2 (ru) 2018-09-06
CN107293287A (zh) 2017-10-24
ES2787894T3 (es) 2020-10-19
KR101884220B1 (ko) 2018-08-01
EP3660845A1 (en) 2020-06-03
CA2940487C (en) 2020-10-27
EP3660845B1 (en) 2022-06-08
JP6493889B2 (ja) 2019-04-03
CN104916292A (zh) 2015-09-16
ES2926360T3 (es) 2022-10-25
ES2787894T9 (es) 2021-12-28
EP3118852A4 (en) 2017-03-29
MX355828B (es) 2018-05-02
KR20180088503A (ko) 2018-08-03
CN107086043A (zh) 2017-08-22
JP2017511901A (ja) 2017-04-27
PT3660845T (pt) 2022-08-01
CA2940487A1 (en) 2015-09-17
US11417353B2 (en) 2022-08-16
EP3118852A1 (en) 2017-01-18
KR102005009B1 (ko) 2019-07-29
RU2016139717A (ru) 2018-04-12
KR20160120764A (ko) 2016-10-18
US20200312353A1 (en) 2020-10-01
EP3118852B1 (en) 2020-02-12
MX2016011750A (es) 2016-12-12
US20190279657A1 (en) 2019-09-12
CN107086043B (zh) 2020-09-08

Similar Documents

Publication Publication Date Title
WO2015135344A1 (zh) 检测音频信号的方法和装置
US10867620B2 (en) Sibilance detection and mitigation
CN104637489B (zh) 声音信号处理的方法和装置
CN106486131A (zh) 一种语音去噪的方法及装置
Tabibian et al. Speech enhancement using a wavelet thresholding method based on symmetric Kullback–Leibler divergence
US20140244247A1 (en) Keyboard typing detection and suppression
US11610601B2 (en) Method and apparatus for determining speech presence probability and electronic device
CN105489226A (zh) 一种用于拾音器的多窗谱估计的维纳滤波语音增强方法
EP3261089B1 (en) Sibilance detection and mitigation
CN114827833B (zh) 啸叫抑制方法、装置、芯片及电子设备
US20180108345A1 (en) Device and method for audio frame processing
CN116312592A (zh) 语音广播扩声系统的啸叫处理方法和装置
Wang et al. Analysis and low-power hardware implementation of a noise reduction algorithm
HK1229058A1 (zh) 检测音频信号的方法和装置
CN114974287A (zh) 降风噪方法、装置、终端设备及存储介质
Peng et al. A Gain Bounded Speech-enhancement Algorithm for Improving Intelligibility

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14885786

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2940487

Country of ref document: CA

ENP Entry into the national phase

Ref document number: 2014386442

Country of ref document: AU

Date of ref document: 20141201

Kind code of ref document: A

ENP Entry into the national phase

Ref document number: 2016556770

Country of ref document: JP

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: MX/A/2016/011750

Country of ref document: MX

ENP Entry into the national phase

Ref document number: 20167025280

Country of ref document: KR

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

REEP Request for entry into the european phase

Ref document number: 2014885786

Country of ref document: EP

WWE Wipo information: entry into national phase

Ref document number: 2014885786

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2016139717

Country of ref document: RU

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: IDP00201606868

Country of ref document: ID

REG Reference to national code

Ref country code: BR

Ref legal event code: B01A

Ref document number: 112016019692

Country of ref document: BR

ENP Entry into the national phase

Ref document number: 112016019692

Country of ref document: BR

Kind code of ref document: A2

Effective date: 20160825