TWI500024B - Sound wave identification system and its method - Google Patents

Description

Acoustic wave identification system and method thereof

The present invention relates to an identification system and method thereof, and more particularly to an acoustic wave identification system and method thereof.

In the physical sense, sound is produced by the propagation of vibrations of solids, liquids, or gases, especially the complex frequency bands of sound vibrations that can be felt by the human ear. The human auditory frequency range is limited to about 20 Hz (Hz). ) to 20 kHz (KHz), and the upper limit generally decreases with age. Other species have different ranges of hearing. For example, some breeds feel a shock to 60,000 Hz. Sound is used by many species to detect danger, navigation, predation, and communication. For example, bats, whales, and dolphins use sound as a basis for navigation, that is, sonar, which is now used for submarine navigation. The sound needs to be transmitted through the medium, so the sound cannot be transmitted in a vacuum environment. Psychoacoustics is the psychological response of the researcher to the auditory, and it is also the psychological response of the average human ear when listening to the sound within the range of 20Hz to 20kHz. However, the average young person can hear about 18kHz. Ears, but as you age, the sensitivity of the human ear to high-frequency sounds decreases, and people who are often harassed by noise or who are used to the music of loud earphones are sensitive to high-frequency sound. It will also drop over time. Most of today's voice recognition applications are voice recognition, which is used to input voice content into a computer system instead of a traditional input device (eg, mouse, keyboard), or as a phone answer content, due to human oral habits, Everyone has a slight difference in the loudness and frequency of each bite when speaking, and there are also obvious differences. Therefore, the accuracy of speech recognition needs further development.

In addition, today's sonic identification technology can also be applied to the use of item identification, as disclosed in the "Coin Identification Device" of Taiwan Patent No. M373528, which utilizes magnetic force. The circle sensing matches the sound of the coin impact to identify the authenticity of the coin, and the technique of sound recognition is as shown in the first figure. Please refer to the first figure, which is a flow chart of conventional sound recognition. As shown in the figure, the conventional coin identification device first uses the sound pickup device to receive the sound when the money is passed through the money channel, and the sound is generated by the inertial impact on the impact zone of the impact bar; the connection is continued as shown in step S20. The sound is an electrical signal; as shown in step S30, the waveform of the electrical signal is adjusted via an amplifier, a filter, and a shaper on the circuit board; as shown in step S40, the frequency, amplitude, and wavelength of the sound according to the coin are displayed. The authenticity judgment is performed by using a microprocessor to compare the parameter value of the sound struck by the coin with the parameter value of the sound struck by the pre-stored genuine coin, and within a certain error tolerance range, the micro-processing The instrument judges the authenticity of the coin by the parameter value of the impact sound. However, when the counterfeit currency is extremely similar to the real coin, the impact sound of the coin does not necessarily have a significant difference. In particular, the waveform of the acoustic wave signal is extremely similar. For the conventional sound recognition device, the frequency, amplitude, and wavelength cannot be recognized very simply. The counterfeit currency, so it needs to be matched with the magnetic ring, otherwise it is inevitable that the counterfeit coin can use the coin-operated device.

In view of this, we will propose an acoustic wave recognition system and its method. In addition to avoiding sound masking and misjudging, the present invention can provide more accurate sound wave recognition and reduce the ratio of false positives than conventional voice recognition and article identification. Moreover, the present invention further utilizes psychoacoustics as the core of acoustic wave analysis to provide accurate sound wave identification, and can recognize sound waves with small differences.

An object of the present invention is to provide an acoustic wave identification system and a method thereof, which utilize psychoacoustics as a basis for acoustic signal analysis to accurately divide an acoustic wave signal for identification.

The present invention provides an acoustic wave identification system and a method thereof, wherein the system uses an acoustic wave capturing unit to capture an acoustic wave signal including a plurality of frequency bands for transmission to a filtering module, and filtering the acoustic wave signal according to the frequency bands. The sound wave signal corresponding to the plurality of frequency bands is generated, and the analyzing unit analyzes the sound wave signal corresponding to the frequency bands by using a psychoacoustic principle to generate an acoustic wave analysis data. Since the present invention analyzes according to a psychoacoustic, the The analysis range of the acoustic analysis data contains 20 Hz to 20 kHz, which is the range of the current accurate acoustic analysis, so the identification unit can recognize the sound by analyzing the acoustic analysis data with accurate range analysis. The source of the wave signal. Thus, the present invention can be accurately used to identify articles, recognize mechanical operating conditions, and recognize speech.

In order to give the reviewer a better understanding and understanding of the structure of the present invention and the efficacies achieved, please refer to the preferred embodiment diagram and the detailed description to illustrate:

80‧‧‧Sonic Identification System

82‧‧‧Sonic extraction unit

84‧‧‧Transition module

842‧‧‧Transfer unit

86‧‧‧Filter module

862‧‧‧ filter

88‧‧‧Analysis unit

90‧‧‧ Identification unit

92‧‧‧Database

The first figure is a flow chart of a conventional sound recognition; the second figure is a block diagram of a preferred embodiment of the present invention; the third figure is a flow chart of a preferred embodiment of the present invention; A schematic diagram of a conversion filtering step of a preferred embodiment; and a fifth diagram of a flow of analysis of a preferred embodiment of the present invention.

Please refer to the second figure, which is a block diagram of a preferred embodiment of the present invention. As shown in the figure, the acoustic wave recognition system 80 of the present invention comprises an acoustic wave capture unit 82, a conversion module 84, a filter module 86, an analysis unit 88 and an identification unit 90. The sound wave capturing unit 82 extracts a sound wave signal from a sound source, wherein the sound wave signal has a plurality of frequency bands, for example, a click sound of the cloth; and the conversion module 84 receives the sound wave signal captured by the sound wave capturing unit 82, and The sound wave signal is converted from the time domain to the frequency domain; the filter module 86 filters the sound wave signal in the frequency domain according to the frequency bands, wherein the filter module 86 corresponds to a psychoacoustic, so the filter module 86 of the embodiment includes 24 filters 842, corresponding to 24 frequency bands analyzed by the human ear, the frequency bands are distributed between 20 Hz and 20 kHz, but can also be reduced to 600 Hz to 16 kHz, which is an effective hearing range for ordinary people. The analyzing unit 88 analyzes the sound wave signal according to the psychoacoustic. Since the analyzing unit 88 corresponds to the psychoacoustic, the analyzing unit 88 analyzes the sound wave signal for each of the 24 frequency bands, that is, the analyzing unit 88 filters the module 86 for 20 The filtering results of Hertz to 20 kHz are analyzed one by one, and a sound wave analysis data is generated.

In response to the above, the identification unit 90 receives the analysis unit 88 for the sound signal. The obtained acoustic wave analysis data is analyzed, and the acoustic wave signal is analyzed according to the acoustic wave comparison data, wherein the identification unit 90 uses the neural network for comparison, and the data is compared by using the neural network. It is a skill that is nowadays so it will not be repeated here. In addition, the acoustic wave recognition system 80 of the present invention further includes a data library 92 for storing the sound wave reference data for the identification unit 90 to read the sound wave reference data for comparing the sound wave analysis data generated by the analysis unit 88. The sound wave comparison data is obtained by the sound wave capturing unit 82 for comparing the sound or voice emitted by the object by the sound wave capturing unit 82, and then converted by the conversion module 84 and the filtering module 86. Filtering, the sound wave comparison data is analyzed by the analyzing unit 88 and stored in the database 92 for identifying the sound source, for example, identifying the cloth to confirm the authenticity through the sound wave analysis data of the cloth. The identification unit 90 can further profile the acoustic analysis data to the database 92 for subsequent more accurate identification. In addition, the acoustic wave recognition system 80 of the present invention can be applied to a computer system, and can be disposed on a single chip, for example, a Field Programmable Gate Array (FPGA).

Please refer to the third drawing, which is a block diagram of a preferred embodiment of the present invention. As shown in the figure, the acoustic wave identification method of the present invention comprises: step S100: capturing an acoustic wave signal; step S200: converting and filtering the acoustic wave signal; and step S300; analyzing the acoustic wave signal corresponding to the frequency bands to generate an acoustic wave analysis And the step S400: analyzing the sound wave according to the sound wave comparison data to identify the sound wave signal.

In step S100, the sound wave signal of the sound source is captured, which is X(n), for example: cloth sound, article or mechanical operation or voice sound wave signal; in step S200, step S100 is taken. The sound wave signal is converted, and the converted sound wave signal is filtered to correspond to the 24 frequency bands of psychoacoustics. As shown in the third figure, it is the signal format of the sound wave signal captured by the conversion step S100. The conversion algorithm of the embodiment is a Fast Fourier Transformation (FFT) for performing acoustic signal conversion, such as the conversion module 84 shown in the fourth figure, which is respectively for 24 frequency bands. Perform a fast Fourier transform. Assuming a discrete Fourier transform on a discrete time domain signal X ( n ), we can get Equation 1 below:

Where ω is the angular frequency and X ( e ^jω ) is the frequency domain function after the conversion, but the invention is not limited thereto, and other time domain domain frequency domain algorithms may be utilized, such as: Laplace conversion or Z conversion; In step S300, the sound wave signal converted by the step S200 is filtered according to the 24 frequency bands corresponding to the psychoacoustic, such as the filter module 86 and the filtering unit 862 shown in the fourth figure. The 24-band band conversion formula is as follows: Equation 2:

Band B ( f ) is mainly a frequency band distinguished by the auditory perception of the human ear and is used to calculate values such as loudness, sharpness and roughness. In the processing of the filtered acoustic wave signal, a bandpass filter is established for the frequency band in which the frequency band B ( f ) is divided into 1/3 octave frequency band, and the bandpass filter is converted. The function H ( s ) can represent Equation 3 as follows:

s is the Laplace operator, Q _∞ is the band factor, and the band corresponding to each 1/3 octave is expressed as shown in Equation 4 below:

f _li ~ f _ui means each frequency corresponding to each of the above filters, which are respectively used for filtering for 24 unequal frequency bands.

In step S300, the signal value X ( f ) corresponding to the acoustic signals of the frequency bands obtained in step S200 is analyzed. As shown in the fifth figure, step S300 further includes: step S310: the loudness extraction Step S320: performing a critical band operation and performing a shadow effect evaluation operation; and step S330: generating sound wave analysis data.

In step S310, the spectrum is divided into Fourier transforms first, and we can calculate and extract the spectral energy value according to the spectrum obtained in step S200, and the equation is as follows: Equation 5: Y ( f )=10. Log ₁₀ {| X ( f )| ² } (5)

The sound pressure conversion of the loudness energy value can be obtained by Equation 6 below, which is as follows:

The loudness energy value L ( f ) can be obtained from Equation 6, and L ( f ) can be substituted into Equation 7 below to find the sum of the energy in each frequency band for each loudness energy value L ( f ) for different frequency bands. The equation 7 is as follows:

Each frequency (f _li ~ f _ui), wherein the series of energy L (i) corresponding to each band B (f) of the energy value of Y (f), i.e. means for each frequency range within each of B (f) The energy contained in a frequency is summed, so the corresponding energy level L ( i ) is calculated for each unequal frequency band B ( f ), so that the energy level L ( i ) included in each frequency band can be expressed as a human ear. The degree of acoustic stress that can be felt in each energy band, and the energy level L ( i ) is psychoacoustically referred to as the Excitation Level.

In step S320, it is calculated according to the loudness energy level obtained in step S310 to obtain the specific loudness value of each frequency band.

First, the formula of the minimum volume threshold information ( LTq ) is obtained by Equation 8, which is as follows:

The calculation formula of the specific loudness is to calculate the critical band shading amount calculation equation L _E by calculating the critical band shading value corresponding to each energy level L ( i ), wherein the critical band shading calculation equation L _E is now known. The shadowing effect evaluates the equation of operation, so it will not be repeated here. Among them, Sones in Equation 9 is the unit of scale for loudness, and Bark is the unit of critical band for Buck. In the minimum volume threshold information (LTQ) than the human ear can hear, the Equation 9 to Equation 8 obtained from the minimum volume threshold information (LTQ) and critical band summed L _E is substituted into equation can be obtained than the loudness calculation Sub N' , the equation 9 is as follows:

In step S330, according to the loudness operator N' corresponding to the 24-band obtained in step S320, the operation is performed to generate sound wave analysis data, and the sound wave analysis data of the embodiment includes the specific loudness value L and the sharp value S. And the roughness value R, wherein the specific loudness total value L is obtained by the following program 10, which is integrated by the loudness operator N' obtained in step S320, which is as follows:

Where L is the total loudness value, which is used as the result of the loudness identification.

After obtaining the loudness operator N' and the specific loudness value L corresponding to each frequency band, the weighting function g ( z ) corresponding to each loudness operator N' can be obtained by a lookup table and weighted, and the weighted The loudness and the division are performed, whereby the sharpness S is obtained according to the equation 12, and the look-up table corresponding to the weighting function g ( z ) is a conventional technique, so it will not be described here. The equation 12 is as follows:

Where S is the sharpness, which is used as the identification value of the acoustic sharpness.

In addition, in this embodiment, Roughness R is obtained by substituting Equation 13 into the loudness operator N' , and the equation 13 is as follows: The amount of change in energy is

It can be seen from Equation 14 that Equation 13 calculates the value of the roughness R from the relationship between the amount of energy change between each loudness operator N' , and Δ f _B is the critical band interval value for each corresponding frequency band, rough The value of the degree R is used to identify the roughness of the sound wave.

In step S400, the sound wave analysis data obtained in step S330 is used for identification, wherein the identification mode is based on the sound wave comparison data stored in the data library 92, and the specific loudness value L and sharpness included in the sound wave analysis data are compared. a value S and a rough value R, which are identified by a neural network based on approximate approximations of the approximation loudness L, sharpness S, and roughness R to identify the source of the acoustic signal captured in step S100, by the above-described specific loudness The value L, the sharp value S and the rough value R are compared to the result to know the state of the sound source, so the invention can be used for item identification, mechanical operation state and voice, wherein the item identification can be used for identifying the cloth, the authenticity of the coin, etc. The mechanical operating state can be applied, for example, to identify engine speeds, voice applications such as the use of voice as an access control, and the use of voice instead of a switch control key.

As can be seen from the above, the above embodiment utilizes psychoacoustic specific loudness value L, sharp value S and rough value R for identification. In addition, psychoacoustic includes Time-Varying Loudness and Tonality. Fluctuation strength, therefore, the present invention can further utilize the variation of loudness, pitch, and audio fluctuation as the basis for identification. The time-varying loudness is the loudness that changes with time, that is, the loudness under transient conditions, and the time-varying loudness is obtained in the same manner as the above-mentioned specific loudness, and the pitch is the human ear's response to the reaction tonal sensation according to different frequency bands. The difference in tone, the amount of audio fluctuations is the amount of change in the response of the audio over time. In addition, the acoustic wave control data of the present invention uses the steps S100 to S300 to capture a control sound wave signal for the control object or the voice before performing the method of the present invention to obtain the sound wave control data and store it in the data library 92.

In summary, the present invention is an acoustic wave recognition system and method thereof, which mainly utilizes a sound extraction unit to capture an acoustic wave signal and transmit it to a filter module to filter sound wave signals corresponding to different frequency bands for use. Corresponding to the acoustic signals of different frequency bands for analysis, and obtaining acoustic wave analysis data, and comparing the acoustic wave analysis data according to the sound wave comparison data, and identifying items such as cloth, wood, metal, or identifying mechanical running sound or voice. .

While the present invention has been described in its preferred embodiments, the present invention is not intended to limit the invention, and the present invention may be modified and modified without departing from the spirit and scope of the invention. The scope of protection is subject to the definition of the scope of the patent application.

Claims (7)

An acoustic wave recognition system comprising: an acoustic wave capturing unit for extracting a sound wave signal; an analyzing unit for analyzing the sound wave signal according to a psychoacoustic analysis to generate a sound wave analyzing data; and an identifying unit for comparing the sound wave according to a sound wave The sound wave analysis data is used to identify the sound wave signal, and the sound wave analysis data includes a Specific Loudness, a Sharpness, and a Roughness; wherein the analyzing unit is configured according to the sound wave signal a plurality of loudness operators are obtained, and the loudness is obtained according to the loudness operators, and the analyzing unit obtains the sharpness according to the specific loudness, and obtains the roughness according to the plurality of frequency bands of the loudness operator and the sound signal. degree. The sound wave identification system of claim 1, further comprising: a database for storing the sound wave control data. The acoustic wave identification system of claim 1, further comprising: a conversion module, receiving the sound wave signal captured by the sound wave capturing unit, and the sound wave signal according to the plurality of frequency bands of the sound wave signal Converting from the time domain to the frequency domain; and a filtering module that filters the frequency bands of the converted acoustic signals for transmission to the analysis unit. The acoustic wave identification system of claim 1, wherein the sound wave signal has a plurality of frequency bands, and the analyzing unit extracts a plurality of energy values corresponding to the frequency bands from the sound wave signal, and performs one according to the energy values. The loudness operation and a critical frequency band operation are used to obtain a complex loudness operator to generate the sound wave analysis data according to the loudness operators. The acoustic wave identification system of claim 1, wherein the identification unit uses a type of neural network to analyze data according to the sound wave comparison data. An acoustic wave identification method, comprising: extracting a sound wave signal; analyzing the sound wave signal according to a psychoacoustic, first extracting a plurality of energy values from the sound wave signal, wherein the energy values correspond to a plurality of frequency bands of the sound wave signal, according to corresponding Some of these bands The magnitude is calculated by the loudness, and then the critical band operation is performed according to the result of the loudness operation, and the complex loudness operator is obtained according to the result of the critical band operation, and the sound wave analysis data is generated according to the loudness operator; wherein the sound wave analysis data further includes A specific loudness is obtained by using the loudness operators, and a sharpness is obtained according to the specific loudness, and a roughness is obtained according to the plurality of frequency bands of the loudness operator and the sound wave signal (Roughness) And; analyzing the data according to the sound wave comparison data to identify the sound wave signal. The method for identifying a sound wave according to claim 6 further comprises: converting the sound wave signal from the time domain to the frequency domain; and filtering the sound wave signal according to the plurality of frequency bands of the sound wave signal for analysis.