JPH1091184A - Sound detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent a state of sound from continuing as a judgment by a sound detection in such a case as noises are suddenly increased in the surroundings and the noise level is maintained. SOLUTION: A speech detection circuit 4 compares a power level of an input signal Sin with a level of an adaptive threshold value ST set by an adaptive threshold value control circuit 3, and judges the state as sound when the level of the input signal Sin is higher than that of the threshold value, and judges as silence when the input is lower than the threshold value. Further, the sound detection circuit 4 is provided with a timer circuit 5 for measuring a succeeding time of sound after the sound detection circuit has judged the input signal as sound. The timer circuit 5 outputs a command signal 6 to an electric power calculation circuit 2 when a judgment signal C has become '1' after having continued for a fixed time T0 or longer, and commands the circuit to multiply the calculated mean electric power P of noises by k(>1.0) for outputting it.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice detection device for detecting a voice signal based on a power level difference between a noise signal and a voice signal included in an input signal.

[0002]

2. Description of the Related Art A voice detector is used for a voice processing device such as an echo canceller of a TV conference system and a voice codec (encoding / decoding device) for voice communication such as a telephone. It has a function of detecting an audio signal from an input signal including the signal.

In detecting such a voice signal, based on only whether the power level of the input signal is higher or lower than a predetermined threshold, the input signal contains voice (voice) or It is possible to determine whether there is only noise (silence). However, the power level of the noise signal changes depending on whether the voice processing device is installed indoors or outdoors, and even when used indoors, whether an air conditioner or the like is used there. The magnitude of the noise will be different. Therefore, in the voice detection method using a fixed threshold, when the noise level at the installation location of the device changes, there is a disadvantage that the noise is erroneously detected as a voice.

In order to prevent such inconvenience, the noise level around the voice processing device is calculated,
An adaptive thresholding method is effective in which a threshold is set to a level always higher than the noise level by a certain value, and a signal above the threshold level is detected.

FIG. 4 shows an example of a conventional voice detection apparatus using the adaptive threshold method.

In FIG. 1, reference numeral 1 denotes a microphone to which a voice signal and a noise signal are input, and 2 denotes a power calculation circuit.
The power calculation circuit 2 calculates the short-time average power P (moving average) immediately before the input signal S _in and outputs the calculated power P to the threshold value calculation circuit 3. Threshold calculation circuit 3 sets an adaptive threshold S _T obtained by adding a constant value to the average power P, and outputs the voice detection circuit 4. The voice detection circuit 4 receives the input signal S _in
From the input signal S _in
Yes and the power level is compared with the level of the adaptive threshold S _T, which is set by the threshold value calculation circuit 3, the higher the better the leveling Le threshold S _T input signal S _in the level of It is determined to be sound, and if low, it is determined to be silent. By this determination, the voice detection circuit 4 outputs a voice / non-voice determination signal C. The sound / non-sound determination signal C is output from the power calculation circuit 2.
And also input to the threshold calculation circuit 3, when the sound is to stop the updating operation of the average power P calculated adaptive threshold S _T of.

FIG. 5 is a signal waveform diagram for explaining the operation of the voice detection device.

The sound shown in FIG. 1A is collected by the microphone 1 together with the gradually increasing ambient noise shown in FIG. 1B, and becomes an input signal S _in .

[0009] FIG. (C) shows the average power P indicated threshold S _T These voice and noise is calculated from the input signal S _in, which is synthesized. The power calculation circuit 2 calculates the average power value of the input signal S _in only the silent section based on the sound / non-speech determination signal C, and the short-time average power P is supplied to the threshold calculation circuit 3. Is output. Adaptive threshold S _T is controlled to a high leveling Le than the average power P by the difference [Delta] P, is output to the voice detection circuit 4.

FIG. 1D shows the waveform of the sound / non-speech determination signal C output from the voice detection circuit 4. The speech detection circuit 4, voice and determine if the level of the input signal S _in is higher than the adaptive threshold S _T, is lower than the adaptive threshold S _T, it is determined that the silence. And this sound /
The silence determination signal C is a signal output of a logical level 1 (hereinafter simply referred to as “1”) and a logical level 0 (hereinafter simply referred to as “0”).

[0011] However, the timing of changing from voiced to silence certain time, wherein the first time is continuously silence only example T _H, the calculation of the average power P of the ambient noise is resumed, sound of the voice detecting circuit 4 / The silence determination signal C becomes “0”. The power calculation circuit 2 time T _H which is set to is a so-called hangover period, for example telephone voice levels for a busy like temporarily reduced, the threshold value S
Even if as below _T, only during the time T _H is a function of maintaining the output of the voice detection circuit 4 to "1".

FIG. 6 is a signal waveform diagram for explaining the operation when the ambient noise increases sharply.

As shown in the figure, for example, the operation of the air conditioner in the room is started, or the door which has been closed until then is opened and noise from the outside enters, and the input signal S _in suddenly rises at a certain point in time. changes, beyond the difference ΔP to the average power P of the amount of change is adaptive threshold S _T, there is a case where the level of the input signal S _in is higher than the adaptive threshold S _T. At this time, the sound / non-speech determination signal C output from the voice detection circuit 4 changes from “0” to “1”. At this point, the calculation of the average power P is stopped at the power calculation circuit 2, also updates the threshold S _T is stopped. In this way only the noise level is high, and the adaptive threshold S _T can not update,
Even though only the noise component is included _{in the} input signal S _in , the sound / non-speech determination signal C does not return to the “0” level forever, and the sound determination continues.

[0014]

As described above, in the conventional voice detection apparatus using the adaptive threshold method, when the surrounding noise level suddenly becomes higher as exceeding the threshold value,
The sound / non-speech determination signal C which was “0” until then becomes “1”
And it is determined that there is sound. Subsequently, upon determining that voice, to thereafter unchangeable threshold S _T of the audio sensing device, when the same noise level is maintained, there is a problem that sound determination continues.

SUMMARY OF THE INVENTION The present invention has been made to solve such a problem, and when ambient noise increases rapidly and its noise level is maintained, a sound state is determined in the determination of voice detection. It is an object of the present invention to provide a voice detection device that prevents continuation.

[0016]

According to a first aspect of the present invention, there is provided a voice detecting apparatus for detecting a voice signal based on a power level difference between a noise signal and a voice signal included in the input signal. Adaptive threshold control means for setting a threshold value for detecting the audio signal from the power level of the audio signal, and audio detection means for comparing the threshold value with the input signal to determine presence / absence of sound And a timer for measuring the time continuously determined to be sound by the voice detection means, and the adaptive threshold control means sets the threshold to the threshold while the sound is determined to be silent. Set to a value corresponding to the power level of the input signal, and when the sound is detected for more than a predetermined time in the timing means,
The threshold value is increased by a predetermined amount.

The adaptive threshold value control means of the voice detection device according to claim 2 repeats the threshold value by a predetermined amount each time the sound is detected for a time exceeding a time preset in the time counting means. Is to increase.

According to a third aspect of the present invention, the adaptive threshold value control means of the voice detection device does not set the threshold value to a predetermined upper limit value or more.

[0019]

BRIEF DESCRIPTION OF THE DRAWINGS FIG.
An embodiment of the present invention will be described.

FIG. 1 is a block diagram showing an example of a voice detection device according to the present invention. The voice detection circuit 4 receives the input signal S _in
From the input signal S _in
Yes and the power level is compared with the level of the adaptive threshold S _T, which is set by the threshold value calculation circuit 3, the higher the better the leveling Le threshold S _T input signal S _in the level of It is determined to be sound, and if low, it is determined to be silent.

The voice detecting device includes a timer circuit 5 for measuring the duration of the voice after the voice detection circuit 4 determines that the voice is voiced.
It has. The timer circuit 5 outputs a command signal 6 to the power calculation circuit 2 when the sound / non-speech determination signal C continuously becomes “1” for a predetermined time (T ₀ ) or more, and the timer signal 5 is already calculated. The average power P of the noise is multiplied by k (> 1.0) and output. In FIG. 1, blocks and signals denoted by the same reference numerals as those of the conventional device in FIG. 4 indicate the same or corresponding blocks and signals. Further, the threshold control means 7 indicates a circuit configuration including the power calculation circuit 2 and the threshold calculation circuit 3.

Next, the operation of the voice detection device shown in FIG. 1 will be described.

In the figure, a sound / non-speech determination signal C is input to a power calculation circuit 2 and a threshold value calculation circuit 3, and when it is determined that there is sound, the calculation of the average power P is stopped. , The average power P of the noise during silence is continuously output to the threshold value calculation circuit 3. As a result, operation of updating the adaptive threshold S _T, will be stopped in case of voiced.

Here, the command signal 6 from the newly added timer circuit 5 is used when the presence / absence of the sound / non-speech determination signal C becomes "1" for a certain time (T ₀ ) or more. The calculation circuit 2 is controlled so as to multiply the average power P of the noise immediately before the sound is determined to be sound and output the result to the threshold value calculation circuit 3. Thus, the average power P output from the power calculation circuit 2 to the threshold calculation circuit 3, after a predetermined time T ₀ has elapsed since it is determined that the active interval, the new value k × P
Is output as

A sound output _Sout is output together with a sound / non-sound determination signal C from the sound detection circuit 4 to a sound processing device (not shown) at a subsequent stage in a sound period as necessary.

Further, the operation of the voice detecting device will be described with reference to a flowchart shown in FIG.

The voice and background noise input from the microphone 1 are input to the power calculation circuit 2 and the voice detection circuit 4 as an input signal S _in , and the voice detection circuit 4 performs a sound / silence determination (step). ST1). When a large input signal S _in is detected than the threshold S _T voice detection circuit 4,
The sound / non-sound determination signal C becomes "1" and is determined to be a sound (step ST2), the process proceeds to step ST3, the timer circuit 5 starts a time counting operation, and the hangover time T
_H is set (step ST4).

Further, in the voice detection circuit 4, the input signal S
When _in is determined that the threshold S _T is less than, the flow proceeds to step ST5, it determines hangover time T _H is whether or not the time over is performed. The hangover time T _H is time-over (T _H ≦ 0) Sometimes, the voice / silence decision signal C "0" (step ST6), the average power P is calculated by the power calculation circuit 2 ( Step ST7). Hangover time T _H is not time-over (T _H> 0) times, speech / silence decision signal C to "1" (step ST8), by reducing the hangover time T _H by the unit time [Delta] T _H (step S
T9), and return to step ST1.

In the timer circuit 5, a constant time T ₀ is set at every Δt.
Is determined (step ST10), and if not, the process returns to step ST1. Then, the above-mentioned sound / non-sound determination is performed, and until the timer circuit 5 determines that the predetermined time T ₀ has elapsed, the process proceeds to step ST10
Thereafter, steps ST1 to ST4 are repeatedly executed. However, S _in <S _T continues hangover time T _H or more voice detection circuit 4 during this time, if voice / silence decision signal C which is the output to "0", the flow proceeds to step ST7, power calculation circuit The calculation of the average power P in 2 is restarted.

In step ST10, the timer circuit 5
When the predetermined time T ₀ has elapsed, the command signal 6 is output from the timer circuit 5 to the power calculation circuit 2. As a result, the average power P calculated by the power calculation circuit 2 is multiplied by k and output (step ST11). As the average power P, a value calculated immediately before the sound is determined to be present in step ST7 is used.

As a result, the value of the new average power P output to the threshold value calculation circuit 3 increases,
In the threshold S _T is updated with the new average power P (step ST12). After that, the timer circuit 5 is reset (step ST13), and the process returns to step ST1 and the sound detection circuit 4 performs the sound / non-speech determination.

[0032] Thus, the threshold S _T, which is updated based on the new average power P is set, since the updated threshold S _T is input to the voice detection circuit 4, the speech detection circuit 4 The sound / non-speech determination is performed based on a different reference from the previous time. Then, until a predetermined time has elapsed T ₀ as before, based on the updated threshold value S _T, it is repeatedly executed further sound interval end judgment.

[0033] Note that after it is determined that the silence in step ST1, the calculation of the average power P of the power calculation circuit 2 in step ST7 is restarted, updates the threshold S _T based on the new average power P in step ST12 (Step ST1)
2) Reset the timer circuit 5 (step ST1)
3) Returning to step ST1, the sound / non-speech determination is repeated.

FIG. 3 is a signal waveform diagram showing a specific example of a series of operations of the above-described voice detection device.

In the figure, the operation from time t ₁ when the sound / non-speech determination signal C in the sound detection circuit 4 changes from “0” (silence) to “1” (sound) is the same as that of the conventional sound detection device. The operation is exactly the same as that shown in FIG. In the voice detection circuit 4 according to the present invention, when the presence / absence determination signal C becomes “1”, the timer circuit 5 starts the timekeeping operation. Until a certain time set in the timer circuit 5 (T ₀₎ has elapsed, the adaptive threshold S _T and the average power P maintains a constant level.

At time t ₂ , the power calculating circuit 2 calculates a value (k × P) obtained by multiplying the average power P calculated last by k (> 1.0) as a new average power and sets the threshold calculating circuit 3
, And the adaptive threshold value _ST is similarly updated to the increased value. This updated threshold _ST is
It becomes higher than the level of the input signal S _in as shown at time t ₃ when the hangover time T _H has elapsed, voice / silence decision signal C of the voice detecting circuit 4 is changed from "1" to "0". Therefore, updating of the average power P is resumed corresponding to the input signal in the power calculation circuit 2 S _in the updating of the adaptive threshold S _T is restarted accordingly.

[0037] In the example shown in FIG. 3, continues to be determined that voiced by ambient noise, the calculated power value by multiplying once k, adaptive threshold S _T is higher than the level of the input signal S _in Such ambient noise is assumed. However, by setting the noise magnitude and constant k, it can be considered as the threshold S _T is not higher than the level of the input signal S _in with only a single multiplication operation. In such a case, as shown in the flowchart of FIG. 2, it will be gradually increased gradually calculated power value every time the predetermined time T ₀ has elapsed.

It is to be noted that, even when speech is normally spoken by a human, even if speech is continued, breaks and breaks always occur in the middle, so that if the audio signal is observed in detail at the power level, the signal level will be It is in an intermittent state. For this reason, if the fixed time T ₀ in the timer circuit 5 of FIG. 1 is selected to be an appropriate value within a range of, for example, about 1 to 10 seconds, there is little possibility that the threshold value _ST is updated by the audio signal. . The example, once for the threshold S _T is even updated by the audio signal, after which the threshold updating S _T only by again noise if Togirere talk takes place, be returned to the appropriate threshold ST Is possible.

As described above, in the above-described speech detection device, the adaptive threshold value _ST is changed after a certain period of time has elapsed even after a sound is produced. Even if erroneous detection of voice is made, it is possible to reliably determine the presence or absence of sound in response to repeated ambient noise.

By modifying the above-described embodiment, the input signal S
calculated value of the average power P _in, or an upper limit to either the adaptive threshold S _T provided to limit such that their values are not above a certain level is also possible to configure the apparatus to perform voice detection is there. In such a voice detection device, even when there is a continuous input of voice, such as an input signal S _in which a music performance such as BGM is used as a voice signal,
Adaptive threshold S _T from the power values that can take the normal ambient noise
If the upper limit is set high, the voice detection operation can be guaranteed.

Although the power calculation circuit 2 calculates a new power value by multiplying the average power P by a predetermined coefficient k, the power calculation circuit 2 may add a predetermined value ΔP.

Further, in the above-mentioned voice detecting device, the command signal 6 of the timer circuit 5 is a command to the power calculation circuit 2. For example, the command signal 6 of the timer circuit 5 is supplied to the threshold value calculation circuit 3, even after a voice, the value of the average power P is held as it is, or may be configured to control only the adaptive threshold S _T to increase gradually every time a predetermined time has elapsed T ₀ .

Therefore, the expression "increase the threshold value by a predetermined amount" described in claim 1 means "increase at a predetermined ratio", "increase of a predetermined value", and a combination thereof. Should.

[0044]

As described above, the speech detection device of the present invention is configured as described above, so that even when the ambient noise changes rapidly, the threshold value is updated in accordance with the new ambient noise. As a result, it is possible to reliably determine the presence or absence of sound.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an example of a voice detection device according to the present invention.

FIG. 2 is a flowchart illustrating an operation of the voice detection device in FIG. 1;

FIG. 3 is a signal waveform diagram showing a specific example of the operation of the voice detection device in FIG.

FIG. 4 is a block diagram illustrating an example of a conventional voice detection device using an adaptive threshold method.

FIG. 5 is a signal waveform diagram illustrating an operation of the voice detection device in FIG.

FIG. 6 is a signal waveform diagram for explaining an operation in a case where ambient noise increases rapidly.

[Explanation of symbols]

1 microphone, 2 power calculation circuit, 3 threshold value calculation circuit, 4 voice detection circuit, 5 timer circuit,
6 instruction signal 7 the threshold control unit, S _in the input signal, S _T adaptive threshold, the average power of P noise,
C Voice / silence determination signal.

Claims (3)

[Claims] 1. A voice detection device for detecting a voice signal based on a power level difference between a noise signal and a voice signal included in an input signal, wherein the voice signal is detected from a power level of the input signal. Adaptive threshold value control means for setting a threshold value; sound detection means for comparing the threshold value with the input signal to determine sound / non-speech; The adaptive threshold value control means sets the threshold value to a value corresponding to the power level of the input signal while it is determined that the sound is silent. A voice detecting device, wherein when the sound is detected for a time longer than a preset time in a timer, the threshold value is increased by a predetermined amount. 2. The apparatus according to claim 1, wherein said adaptive threshold value control means repeatedly increases said threshold value by a predetermined amount each time said sound is detected for a time exceeding a time preset in said timekeeping means. The voice detection device according to claim 1. 3. The apparatus according to claim 2, wherein said adaptive threshold value control means does not set said threshold value to a predetermined upper limit value or more.