JPH05130067A - Variable threshold level voice detector - Google Patents

Abstract

PURPOSE:To ensure almost equal voice quality regardless of the quantity of the S/N. CONSTITUTION:This detector is provided with a voice signal power calculation circuit 1 calculating a voice signal component in an input signal, a noise power calculation circuit 2 calculating noise power included in the input signal, a relative ratio calculation circuit 3 calculating a relative ratio of a voice signal power to a noise power, a hang-over calculation circuit 4 calculating a hang-over time in response to the relative ratio, a threshold level generating circuit 5 calculating and outputting a discriminated threshold level in response to the noise power, a discrimination circuit 6 discriminating the presence of the voice signal by comparing the input signal with the discrimination threshold level, and a hang-over addition circuit 7 adding a hang-over time to the result of discrimination as the presence of the voice signal when the result of discrimination indicates a change from the presence of the voice signal into the absence of the voice signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a variable threshold type voice which determines the presence or absence of a voice signal by comparing an input signal with a decision threshold set according to noise power and adds hangover to the decision result. Regarding the detector.

[0002]

2. Description of the Related Art Generally, DSI (Digital Speech Inter
Abbreviation for polation. ) Is used to determine if an audio signal is present on the input channel to the device using the technology. Also, since the characteristics of the voice detector have a great influence on the communication quality of the device and further on the system design, the voice detector should respond to the voice signal as quickly as possible, and be sensitive to noise. Therefore, it is required that it should not respond as much as possible, that it can be realized by simple hardware, and that it does not increase the activity (the time ratio at which the voice detector indicates that there is a voice signal) more than necessary.

Conventionally, as this type of voice detector, for example, INTELSAT (International) is used.
DCME (Digital Circuit Multiplication) specified by Telecommunications Satellite Organization)
Equipment) device determines the presence or absence of a signal by comparing an input signal input at each sample time with a determination threshold value that is set higher as the noise power increases according to the noise power included in this input signal. A variable threshold type voice detector that adds and outputs as a voice signal for a predetermined hangover time (depending on the duration of the signal) is output to the determination result when the determination result changes from the presence of a voice signal to the absence of a voice signal. Has been.

This voice detector, as shown in FIG.
It is used by being incorporated in the main signal path of a device using SI technology. In FIG. 7, the delay circuit 50 is for adjusting the detection time of the voice detector 40, and the selection circuit 60 uses the delayed input signal which is the output of the delay circuit 40 as the judgment output from the voice detector 40. It is based on the selection. This is because it is possible to improve the line efficiency by selecting and transmitting only audio signals.

FIG. 8 shows an example of input / output waveforms of this variable threshold type voice detector.

For example, when the relative ratio between the voice signal power and the noise power of the input signal 30 (hereinafter referred to as "S / N ratio") is small, the noise is relatively large with respect to the voice signal. A first determination threshold value 31 which is a determination threshold value of a relatively large level shown in 8 is set, and the determination result of the presence or absence of an audio signal, which is the result of the comparison between the input signal 30 and the determination threshold value 31 at this time, is the first determination result. It looks like 33. on the other hand,
When the S / N ratio of the input signal is large, the noise level is relatively low with respect to the voice signal, so the second determination threshold value 32, which is the determination threshold value of the relatively low level shown in FIG. The determination result of the presence / absence of the audio signal is as the second determination result 36. In addition, these first determination results 3
A fixed hangover time th is added to each of the third and second determination results 36, which are output to the selection circuit 60 described above as the first determination output 35 and the second determination output 37.

In FIG. 8, reference numeral 34 indicates a delayed input signal which is the output of the delay circuit 50 input to the selection circuit 60, and reference numeral 38 indicates the output of the selection circuit. Further, tdl in the first determination result 33 and tdh in the second determination result 36 are operation time of the voice detector 40, in other words, time required for detection (hereinafter, referred to as “detection time”),
Tf in the delayed input signal 34 is the delay amount.

[0008]

In the variable threshold type voice detector used in the DCME device defined by the above-mentioned conventional Intelsat specification, the hangover time th added to the voice signal in order not to increase the activity more than necessary. Is specified within 30 msec. Therefore, when the S / N ratio is small and the judgment threshold value for judging the presence / absence of the audio signal is set high like the first judgment threshold value 31 in the figure, as a result, the judgment output which is the input to the selection circuit is First
Therefore, there is an inconvenience that a voice signal component of low power cannot be detected, resulting in dropout (breaking in the middle of a word) or missing word ending (breaking in the end of a word). ..

Further, the detection time td is specified corresponding to the S / N ratio as shown in FIG. 9 in the DCME device specified by the use of Intelsat. Therefore, the delay amount tf of the delay circuit is set as shown in FIG. 8 so that the beginning of the word is not lost (the rising part of the word is cut off) even when the detection time is the maximum, and the delay amount is a constant value. ing. Therefore, when the S / N ratio is large and the detection time is short, as is apparent from the second determination output 37 and the selection circuit output 38 in FIG. 8, the noise part before the beginning and after the ending is detected as the voice signal. There was also the inconvenience of increasing the activity more than necessary.

[0010]

SUMMARY OF THE INVENTION The object of the present invention is to improve the disadvantages of the prior art, and more particularly to a variable threshold type voice detector capable of ensuring almost the same voice quality without being affected by the magnitude of the S / N ratio. To provide.

[0011]

A variable threshold type voice detector of the present invention comprises a voice signal power calculation circuit for calculating a voice signal component of an input signal input at each sample time, and a noise power included in the input signal. A noise power calculation circuit for calculating, a relative ratio calculation circuit for calculating the relative ratio between the voice signal power and the noise power based on the outputs of the voice signal power calculation circuit and the noise power calculation circuit, and this relative ratio calculation circuit The hangover calculation circuit that calculates the hangover time according to the relative ratio, the threshold value generation circuit that calculates and outputs the judgment threshold value according to the noise power output from the noise power calculation circuit, and the input signal and the threshold value generation circuit A determination circuit that determines the presence or absence of a voice signal by comparing the output determination threshold value with the output of this determination circuit, and when the determination result changes from the presence of a voice signal to the absence thereof, And a hangover adding circuit for adding as there audio signal hangover time duration output from the hangover calculating circuit to a constant result. With such a configuration, the above-described object is to be achieved.

[0012]

[First Embodiment] A first embodiment of the present invention will be described below with reference to FIGS.

FIG. 1 shows the configuration of the voice detector of this embodiment. In the embodiment of FIG. 1, a voice signal power calculation circuit 1 that calculates a voice signal component of an input signal that is input at each sample time, a noise power calculation circuit 2 that calculates noise power included in the input signal, and a voice signal. An S / N ratio calculating circuit 3 as a relative ratio calculating circuit for calculating an S / N ratio which is a relative ratio between the voice signal power and the noise power based on the outputs of the power calculating circuit 1 and the noise power calculating circuit 2; A hangover calculation circuit 4 for calculating a hangover time according to the S / N ratio calculated by the S / N ratio calculation circuit, and a noise power calculation circuit 2
A threshold generation circuit 5 that calculates and outputs a determination threshold value according to the noise power output from the input signal, and a determination circuit that determines the presence or absence of a voice signal by comparing the input signal with the determination threshold value output from the threshold value generation circuit 5. 6 and the hangover for adding the hangover time output from the hangover calculation circuit 4 to the judgment result when the judgment result changes from the presence of the sound signal to the absence of the sound signal. And an additional circuit 7. In FIG. 1, reference numeral 9 indicates an input terminal, and reference numeral 10 indicates an output terminal to the selection circuit 60 described above.

Next, the overall operation of the first embodiment constructed as described above will be explained.

The input signal input from the input terminal 9 is input to the voice signal power calculation circuit 1, the noise power calculation circuit 2 and the determination circuit 6. In the voice signal power calculation circuit 1, when the output of the determination circuit 16 indicates that there is a voice signal, that is, when the input signal is a voice signal, the signal power of the input signal is calculated, and the S / N ratio calculation circuit is calculated as the voice signal power. Output to 3. on the other hand,
The noise power calculation circuit 2 calculates the signal power (noise power) of the input signal when the output of the determination circuit 6 indicates that there is no voice signal, that is, when the input signal is noise, and the S / N ratio calculation circuit as the noise power. 3 and the threshold value generation circuit 5.

In the S / N ratio calculation circuit 3, the S / N ratio (here, the average voice signal power) is calculated from the voice signal power output from the voice signal power calculation circuit 1 and the noise power output from the noise power calculation circuit 2. Is defined as the relative ratio between the average noise power and the average noise power.)
Is output to the hangover calculation circuit 4. The hangover calculation circuit 4 calculates the hangover time according to the S / N ratio output from the S / N ratio calculation circuit 3 (see FIG. 3) and outputs it to the hangover addition circuit 7.

In the threshold generation circuit 5, the determination circuit 6 is based on the noise power output from the noise power calculation circuit 2 described above.
The threshold value (see reference numerals 21 and 22 in FIG. 2) for determining the presence / absence of the audio signal is calculated and output. Here, this determination threshold is set higher as the noise power is higher. The determination circuit 6 determines the presence / absence of an audio signal by comparing the input signal (see reference numeral 20 in FIG. 2) with the determination threshold value output from the threshold value generation circuit 5, and the hangover addition circuit 7 determines the determination result (see FIG. 2 reference numerals 23 and 25) are output. In the hangover addition circuit 7, when the determination result output from the determination circuit 6 changes from the presence of a voice signal to the absence thereof, the hangover time th output from the hangover calculation circuit 4 is set as the presence of the voice signal in the determination result. It is added and output as a determination output (reference numerals 24 and 26 in FIG. 2).
reference). As a result, the selection circuit 60 described above selects the delayed input signal which is the output of the delay circuit 50 based on this determination output.

Now, the relationship between the S / N ratio and the hangover time will be described with reference to FIG. In this figure, reference numeral 20 is an input signal, reference numeral 21 is a first determination threshold value, reference numeral 2
Reference numeral 2 indicates a second determination threshold value, reference numeral 23 indicates a first determination result, reference numeral 24 indicates a first determination output, reference numeral 25 indicates a second determination result, and reference numeral 26 indicates a second determination output. In addition,
Here, the magnitudes of the levels of the first determination threshold and the second determination threshold are set to be opposite to those of the conventional example described above, but there is no special meaning.

For example, when the S / N ratio of the input signal 20 is large, noise is relatively low with respect to the voice signal, and therefore the threshold generation circuit 5 outputs a relatively low level as the determination threshold. When this determination threshold value is the first determination threshold value 21, the determination result indicating the presence or absence of the audio signal output from the determination circuit 6 is the first determination result 23. On the other hand, when the S / N ratio of the input signal 20 is small, the threshold generation circuit 15 outputs a relatively large level as the determination threshold because the noise is relatively large with respect to the voice signal. When the determination threshold is the second determination threshold 22, the determination result indicating the presence or absence of the audio signal output from the determination circuit 6 is the second determination result 25.

Therefore, in order to secure the same voice quality as when the S / N ratio of the input signal 20 is small, the first judgment result 23 and the second judgment result 25 are obtained. As can be seen by comparison, when the S / N ratio is small, S /
When the N ratio is large, it is necessary to add the above hangover time th to the determination result. In this embodiment, as shown in FIG. 3, the hangover calculation circuit 4 calculates a relatively long hangover time when the S / N ratio is small, and a relatively short hangover time when the S / N ratio is large. It is designed to calculate time. For this reason, in the present embodiment, the judgment outputs as shown by the first judgment result 24 and the second judgment result 26 in FIG. 2 are obtained, and substantially the same voices are not affected by the magnitude of the S / N ratio. Quality is secured.

[0021]

[Second Embodiment] Next, a second embodiment of the present invention will be described with reference to FIGS. Here, the same components as those in the first embodiment described above are designated by the same reference numerals, and
The description will be omitted.

In this embodiment, the output of the hangover addition circuit 7 is delayed according to the judgment threshold value calculated by the S / N ratio calculation circuit 3 at the output stage of the hangover addition circuit 7 in the first embodiment. It is characterized in that a delay adjusting circuit 8 for adjusting the detection time is additionally provided. The configuration of the other parts is the same as that of the first embodiment.

In the second embodiment thus constituted, the output of the hangover circuit 7 is delayed by the delay adjusting circuit 8 for the time shown in FIG. 5 corresponding to the S / N ratio as shown in FIG. Therefore, the time obtained by adding the delay time shown in FIG. 5 to the operation time shown in FIG. 9 is substantially the detection time of the voice detector. As a result, as the output of the delay adjustment circuit 8, the adjustment determination output 39 in which the detection time td is corrected by ta is obtained as shown in FIG. 6, and the selection circuit output 38 shown in the figure (this is shown in the conventional example). It is the same waveform as the above.), The selection circuit 60 does not select the noise part before the beginning of the word, and the increase in activity due to the noise part before the beginning of the word can be prevented. Also in the second embodiment, S is the same as in the first embodiment.
Since the hangover time is set according to the / N ratio, as a result, in the present embodiment, substantially the same voice quality can be secured without depending on the S / N ratio, and It is also possible to prevent an unnecessary increase in activity.

[0024]

As described above, according to the present invention, the hangover calculating circuit calculates the hangover time to be added to the result of the determination of the presence or absence of the audio signal, according to the S / N ratio. The over time can be made relatively long when the S / N ratio is small, and can be made relatively short when the S / N ratio is large, whereby almost the same voice is always received without being affected by the S / N ratio. It is possible to provide an unprecedented excellent variable threshold type voice detector capable of ensuring quality.

Further, in the invention described in claim 2, in particular, there is an effect that it is possible to prevent an increase in activity more than necessary in addition to securing an equivalent voice quality which is not affected by the S / N ratio.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a first exemplary embodiment of the present invention.

FIG. 2 is a diagram showing an example of input / output waveforms of respective parts in the embodiment of FIG.

3 is an S set by the hangover calculation circuit of FIG.
It is a diagram which shows the relationship between / N ratio and hangover time.

FIG. 4 is a block diagram showing a configuration of a second exemplary embodiment of the present invention.

5 is a diagram illustrating a delay amount and S / of a delay circuit in the embodiment of FIG.
It is a diagram which shows the relationship with N ratio.

FIG. 6 is a diagram showing an adjustment determination output in the embodiment of FIG. 4 together with a selection circuit output.

FIG. 7 shows a portion of the main signal path of a DCMA device.

FIG. 8 is a diagram showing an example of input / output waveforms of a conventional variable threshold type voice detector.

FIG. 9 is a diagram showing a relationship between an S / N ratio and a detection time of a voice detector in a DCME device defined by Intelsat specifications.

[Explanation of symbols]

 1 Voice Signal Power Calculation Circuit 2 Noise Power Calculation Circuit 3 S / N Ratio Calculation Circuit as Relative Ratio Calculation Circuit 4 Hangover Calculation Circuit 5 Threshold Generation Circuit 6 Judgment Circuit 6 7 Hangover Addition Circuit 8 Delay Adjustment Circuit

Claims (2)

[Claims] 1. A voice signal power calculation circuit for calculating a voice signal component of an input signal input at each sampling time, and a noise power calculation circuit for calculating noise power included in the input signal.
A relative ratio calculation circuit for calculating a relative ratio between the voice signal power and the noise power based on the outputs of the voice signal power calculation circuit and the noise power calculation circuit, and a hang according to the relative ratio calculated by the relative ratio calculation circuit. A hangover calculation circuit that calculates an overtime, a threshold generation circuit that calculates and outputs a determination threshold value according to noise power output from the noise power calculation circuit, and a determination output from the input signal and the threshold generation circuit A judgment circuit for judging the presence or absence of a voice signal by comparing with a threshold value, and when the judgment result changes from the presence of a voice signal to the absence of a sound signal received from the judgment circuit, the judgment result is output from the hangover calculation circuit. A variable threshold type voice detector characterized by comprising a hangover adding circuit for adding a hangover time corresponding to the presence of a voice signal. 2. A delay adjusting circuit for adjusting the detection time by delaying the output of the hangover adding circuit according to the determination threshold value calculated by the relative ratio calculating circuit is provided at the output stage of the hangover adding circuit. 2. The variable threshold type voice detector according to claim 1, wherein