JPH04335177A - Sound detector - Google Patents

Abstract

PURPOSE:To surely and easily perform the identification of the existence of a target sound in mixed sounds. CONSTITUTION:A mixed sound signal v0 in which an objective sound signal a0 and a nonobjective sound signal from a receiver 18 is input to a first module 11 having adaptive filters 13, 15 and a second module 12 receiving the output therefrom. Furthermore a reference signal u0 from a reference signal input part 17 is input to the first module 11, an input signal is adapted and identified in each module for the purpose of estimation and elimination and the objective signal a0 correlated with the reference signal u0 is obtained as the output signal of the second module 12. As the result, the existence of the objective sound signal a0 correlated with the reference signal u0 can be confirmed clearly and in addition the extraction of the objective sound signal a0 can be surely and easily performed.

Description

[Detailed description of the invention]

[Object of the invention]

0002

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound detection device for detecting a target sound in a mixed sound including, for example, a target sound and a non-target sound.

0003

[Background Art] Searching out whether a known target sound is included in the waveform of various sounds. For example, when a certain device is operating among various devices being operated. It is possible to judge whether or not the equipment contains a known specific sound that the equipment has, or to determine whether such abnormal sounds or related sounds are being emitted based on the recorded abnormal sounds. In some cases, it is possible to detect whether a device is present, or to remove only a specific sound and extract the remaining target sounds.

On the other hand, adaptive noise cancellers have conventionally been used as devices that can clearly capture sound from a specific sound source in a noisy environment where various sounds are mixed. This adaptive noise canceller has a sound source sound s from a sound source S and an ambient sound n as shown in FIG.
a microphone 1 that collects a sound, a microphone 2 that is arranged so as to collect only an ambient sound n without collecting a source sound s;
It is configured to include a module 4 having an adaptive filter 3 that processes signals input from each of the microphones 1 and 2. Note that the source sound signal collected by microphone 1 is s0, the ambient sound signal is np, and the ambient sound signal collected by microphone 2 is nr.

In the device configured as described above, the ambient sound signal np input together with the source sound signal s0 is input to the module 4, and the ambient sound signal nr is input to the adaptive filter 3, and adaptive identification is performed while updating the filter coefficient h. It operates to estimate and cancel the ambient sound signal np. That is, if the output signal of the adaptive filter 3 is y0, then
The error signal ε0, which is the output of module 4, is ε0 = s0 + np - y0. Then, the mean square error E(ε
0 2 ) is E(ε0 2 )=E(s0 2 )+2E[s0 (
np −y0 )]+E[(np −y0 )2 ], and since s0, np and y0 are generally considered to be random signals, they can be assumed to be uncorrelated with each other. Therefore, the second term converges to 0 by repeatedly averaging, and finally the mean square error E(ε0
2) is E(ε0 2 )=E(s0 2 )+E[(np −
y0 )2 ], while np −y0 = ε0
-s0, the adaptive filter 3 uses E(ε0
2), the error signal ε0 and the sound source signal s0
The difference between is also minimized, and the error signal ε0 is the source sound signal s0
It is approximated by In this way, the signal s0 of the source sound s from which the ambient sound n has been removed can be obtained as the output of the module 4.

However, this device is intended to clearly obtain the source sound s from a mixture of the source sound s from the sound source S and the ambient sound n; It is not possible to determine whether there is a specific sound in the sound that corresponds to a reference signal obtained in advance, that is, a target sound or a sound that is correlated with the target sound, and the presence of a sound source that emits the target sound. cannot be made clear. In addition, if the target sound obtained in advance is an abnormal sound of the equipment,
It has not been possible to meet the demand for finding out whether there is an abnormality in the equipment using this target sound. Alternatively, to describe the terminology used in the conventional example, there is a requirement to extract only the surrounding sound n excluding the source sound s in the mixed sound, that is, excluding only the specific sound or sounds correlated to the specific sound. The remaining target sounds could not be extracted.

[0007]

[Problem to be solved by the invention] The present invention was made in view of the situation where it is impossible to determine whether or not a target sound exists in a mixed sound as described above. To provide a sound detection device that can reliably and easily determine whether or not there is a reference signal obtained in advance in a sound, or extract only target sounds that are correlated with the reference signal. There is a particular thing.

[Configuration of the invention]

[0009]

[Means for Solving the Problems] The sound detection device of the present invention includes:
A sound receiver that collects a mixed sound containing a mixture of target and non-target sounds, a reference signal input section into which a reference signal is input, a reference signal from this reference signal input section and a mixed sound signal from the sound receiver. a first module that inputs and adaptively identifies a target signal with an adaptive filter, estimates and eliminates a target signal correlated with a reference signal in a mixed sound signal, and outputs a non-target sound signal; A second module receives the output signal and the mixed sound signal from the sound receiver, performs adaptive identification using an adaptive filter, estimates and eliminates non-target signals in the mixed sound signal, and outputs a target sound signal. It is characterized by the fact that
The adaptive filter of the second module is characterized in that it converges earlier than the adaptive filter of the second module. In addition, there are a plurality of sound receivers arranged at a distance to collect a mixture of target sounds and non-target sounds, a reference signal input section into which a reference signal is input, and a reference signal input section corresponding to each sound receiver. A mixed sound signal from the sound receiver and a reference signal from the reference signal input section are input and adaptively identified by an adaptive filter,
a first module that estimates and eliminates a target signal correlated with a reference signal in the mixed sound signal and outputs a non-target sound signal; and a first module that is provided corresponding to each sound receiver and mixes from the sound receiver. a second module that receives the sound signal and the output signal of the first module, performs adaptive identification using an adaptive filter, estimates and eliminates a non-target signal in the mixed sound signal, and outputs a target sound signal;
It is characterized by comprising a third module that performs adaptive identification by changing the filter coefficients of the adaptive filter so that the difference between the output signals inputted from these second modules is minimized, The adaptive filter of the first module converges before the adaptive filter of the second module, and the adaptive filter of the second module converges before the adaptive filter of the third module. The third module is configured to include a filter having a preset filter coefficient.

0010

[Operation] The sound detecting device configured as described above has two functions: one is to detect a mixed sound signal from a sound receiver in which a target sound signal and a non-target sound signal are mixed, and a first module provided with an adaptive filter; This output is input to the second module connected to the input, and the reference signal is input from the reference signal input section to the first module, and each module adaptively identifies, estimates, and cancels the input signal. Then, a target sound signal correlated with the reference signal is obtained as an output signal of the second module. Another method is to input target sound signals with different phases output from a plurality of second modules to a third module provided with an adaptive filter, and adapt the signals so that the difference between the input signals is minimized. The direction of the sound source is determined from the phase difference. Therefore, the existence of a target sound signal that has a correlation with the reference signal in the mixed sound can be clearly identified, and the target sound signal can be extracted reliably and easily. Furthermore, the correct direction of arrival of the target sound can be identified reliably and easily.

[0011]

Embodiments Hereinafter, embodiments of the present invention will be described with reference to the drawings. Prior to describing embodiments, the principle of the present invention will be explained with reference to FIG. In the figure, 11 is a first module, and 12 is a second module. The first module 11 is provided with an adaptive filter 13 having a filter coefficient h1 and an arithmetic unit 14.
is connected so that the input signal u0 of
4 is connected so that the difference between the second input signal v0 and the output signal z1 of the adaptive filter 13 is calculated and output. Furthermore, the adaptive filter 13 uses the output signal e1 of the calculation section 14.
The filter coefficient h1 is updated by =v0-z1, and the output signal e1 of the calculation unit 14 is decreased and adaptively identified so as to converge to the minimum value. That is, in the first module 11, the filter coefficients are updated using LMS (Lea
St Mean Square ) algorithm, etc. is used to converge. h1 new = h1 old
+μ1 e1 u0 Note that μ1 is the change rate (convergence coefficient). Similarly, the second module 12 includes an adaptive filter 15 with a filter coefficient h2 and a calculation unit 1.
6 is connected to the adaptive filter 15 so that the output signal e1 of the calculation section 14 of the first module 11 is input, and the calculation section 16 is connected to receive the second input signal v0 and the output of the adaptive filter 15. It is connected so that the difference with the signal z2 is calculated and output. Also, the adaptive filter 15
is adaptively identified so that the filter coefficient h2 is updated by the output signal e2 = v0 - z2 of the calculation unit 16, and the output signal e2 of the calculation unit 16 decreases and converges to the minimum value. Similarly, in the second module 12, the filter coefficients are updated using the following equation, etc., and convergence is achieved. h
2 new =h2 old +μ2 e2 e1
Note that μ2 is a change rate (convergence coefficient). Furthermore, if the adaptive filters 13 and 15 perform adaptive identification and convergence at the same time, they may interfere with each other and fail to converge. Therefore, the convergence coefficient of the adaptation of the adaptive filter 13 is
The convergence coefficient of adaptation No. 5 is set to be sufficiently larger than the convergence coefficient of adaptation No. 5, or convergence is performed by the adaptive filter 15 after the convergence by the adaptive filter 13 is performed.

[0012] Here, the second input signal v0 is expressed as v0 = a
0+b0, and if a0 and the first input signal u0 are correlated signals, then the first module 11
Then, the output signal z1 of the adaptive filter 13 converges to a0, and the output signal e1 of the arithmetic unit 14 becomes b0. Similarly, in the second module 12, the output signal z2 of the adaptive filter 15 converges to b0, and the output signal e2 of the calculation unit 16
=a0. In this way, by passing the first input signal u0 and the second input signal v0 through the first module 11 and the second module 12, a correlation is generated from the second input signal v0 to the first input signal u0. A certain signal a0 can be finally extracted. The present invention is constructed based on such a principle.

First, a first embodiment of the present invention will be explained with reference to FIG. Note that the same parts as in FIG. 1, which explain the principle of the present invention, are given the same reference numerals and the description thereof will be omitted. FIG. 2 is a schematic configuration diagram, and in the figure, 17 is the first module 1.
A reference signal input unit that inputs the first input signal u0 of 1, that is, a reference signal obtained in advance and whose waveform etc. are known, and whose output is input to the adaptive filter 13 of the first module 11.
is connected so that it is input to the 18 is a receiver that collects the second input signal v0 of the first module 11 and the second module 12, that is, the mixed sound v of the target sound a and the non-target sound b at a predetermined position, and outputs the mixed sound signal v0. The sound generator is connected so that its output is branched and input to calculation units 14 and 16 of the first module 11 and the second module 12, respectively. Reference numeral 19 denotes a determining section connected to input the output of the second module 12. This determining section 19 determines the output signal e2 from the calculating section 16, and correlates it with the reference signal u0 in the mixed sound signal v0. It is determined whether a certain target sound signal a0 exists.

According to the first embodiment configured in this way, for example, one device S among the devices in operation emits the target sound a as an abnormal sound, and other devices emit the non-target sound. Assuming that b is produced, target sound a and non-target sound b
is collected by the sound receiver 18 as a mixed sound v=a+b, and the first
The calculation unit 1 of the module 11 and the second module 12
The signal is input to 4 and 16. On the other hand, in order to detect the presence or absence of equipment failure, a reference signal u0 correlated with abnormal sounds recorded in advance is input to the reference signal input section 17, and the output of the reference signal input section 17 is input to the first module 11. is input to the adaptive filter 13 of. First, adaptive identification progresses in the first module 11 where the convergence coefficient of filter adaptation is sufficiently large, and the mixed sound signal v0=a0 +b0
Among them, the target sound signal a0 that has a correlation with the reference signal u0
The non-target sound signal b0 from which is removed is the first module 1
1 output signal e1=b0.

In the second module 12, the output signal e1 = b0 of the first module 11 is input to the adaptive filter 15, and the adaptive identification is performed by the mixed sound signal v0 = a0 + b0 input to the arithmetic unit 16. The output signal e1 = b0 of the first module 11 is removed from the mixed sound signal v0 = a0 + b0, and the target sound signal a0 is replaced by the output signal e2 = a0 of the second module 12.
is output as Then, the output signal e2 = a0 of the second module 12 is input to the discriminator 19, and the second
Output signal e2 from the calculation unit 16 of the module 12 =
a0 is determined, and it is found that there is a target sound signal a0 correlated with the reference signal u0 in the mixed sound signal v0 = a0 + b0, and it is found that there is an abnormality in the device S.

As described above, according to the first embodiment, even in a situation where a target sound a having a correlation with a known reference signal u0 exists together with a non-target sound b, the target sound a and the non-target sound b can be captured together as a mixed sound signal v0 = a0 + b0, and it can be accurately and reliably known that there is a target sound a correlated with the reference signal u0,
It is easy to determine whether the device S is in operation or whether there is an abnormality in the device S, and it can be determined remotely without approaching the device S.

Next, a second embodiment of the present invention will be explained with reference to FIG. The main part of this embodiment includes the sound receiver and two first and second modules of the first embodiment, and the same parts are given the same reference numerals and the corresponding parts are indicated. For parts where , a dash symbol (
') and the explanation will be omitted. FIG. 3 is a schematic configuration diagram. In the figure, two sound receivers 18, 18' are provided apart from each other, and these sound receivers 18, 18' output mixed sound signals having different phases, v0 = a0 + b0, v0′=a0
'+b0' is output. The sound receivers 18, 18' are
The output mixed sound signals v0, v0' are connected so as to be branched and input to first modules 11, 11' and second modules 12, 12', respectively. A reference signal input section 17 is connected to the first module 11, 11' so that the reference signal u0 is input thereto.

20 is the third module, which includes the second
The output signals e2 and e2' of the arithmetic units 16 and 16' of the modules 12 and 12' are connected to each other. The third module 20 is provided with an adaptive filter 21 having a filter coefficient h3 and an arithmetic section 22, and the adaptive filter 21 is provided with an arithmetic section 16' of the second module 12'.
The output signal e2' of the second module 12 is connected to the arithmetic unit 22, and the difference e3 between the output signal e2 of the arithmetic unit 16 of the second module 12 and the output signal z3 of the adaptive filter 22 is calculated and output. connected so that Furthermore, the adaptive filter 21 receives the output signal e3 of the calculation unit 22.
The filter coefficient h3 is updated by
Adaptive identification is performed so that the output signal e3 of is decreased and converged to the minimum value. The third module 20 outputs the output signal e2' of the second module 12', and the output signals e2, e2 of the second modules 12, 12' obtained from the filter coefficient h3 of the adaptive filter 21.
′ is output.

Reference numeral 23 denotes a determining section to which the output of the third module 20 is connected. The discriminator 23 uses the input output signal e2' of the second module 12' to determine the mixed sound signal v0 = a0 + b0
It is determined whether or not there is a target sound signal a0 correlated with the reference signal u0, and the arrival direction of the signal received by the sound receivers 18 and 18' is determined based on the input phase difference. . Note that the adaptive filter 2 of the third module 20
The convergence coefficient of 1 is set smaller than the convergence coefficient of the adaptive filters 15, 15' of the second modules 12, 12'.

According to the second embodiment configured as described above, the mixed sound signals v0, v0' having different phases outputted from the sound receivers 18, 18' and the reference signal inputted to the reference signal input section 17 u0 is the first module 11, 11'
and the second modules 12 and 12', respectively, and the mixed sound signals v0 = a0 + b0, v0' = a0
Signals a0 and a0' in '+b0' are reference signal u0
, the output signals e2 and e2' of the second modules 12 and 12' are e2 = a
0, e2'=a0'. Since each output signal e2 = a0 , e2 ′ = a0 ′ is a correlated signal, it becomes an output signal z3 through the adaptive filter 21 with the filter coefficient h3 in the third module 20, and the output signal z3 is output in the calculation unit 22. 12 output signals e
2, the output signal z3 of the adaptive filter 21 is subtracted from the signal e3, and the filter coefficient h3 of the adaptive filter 21 is updated by the output signal e3 of the arithmetic unit 22, and the output signal e3 is converged to zero. . Note that the convergence coefficient of the adaptive filter 21 of the third module 20 is the same as that of the adaptive filters 13, 13', 15, 15 of the first module 11, 11' and the second module 12, 12'.
Since the convergence coefficient of the adaptive filter 21 is set smaller than the convergence coefficient of the adaptive filter 13, 13', 15,
It is completed after the convergence of 15' is completed. As a result, the phase difference between the output signals e2 and e2' of the second modules 12 and 12', that is, the phase difference between the target sound signals a0 and a0', is extracted as the filter coefficient h3 of the adaptive filter 21. Also, the target sound signal a is selected from the mixed sound signal v0'=a0'+b0' received by the sound receiver 18'.
Only 0' is output and input to the discriminator 23.

Then, the discriminator 23 finds that there is a target sound signal a0' correlated with the reference signal u0 in the mixed sound v=a+b, and for example, the reference signal u0 is correlated with the abnormal sound of the device S. If so, it is determined that there is an abnormality in the device S, and the direction of arrival of the target sound a can be correctly identified based on the phase difference.

As described above, according to the second embodiment, the same functions and effects as those of the first embodiment can be obtained, and the direction of arrival of the target sound a correlated with the reference signal u0 can be correctly and reliably obtained. It will be done.

Note that the present invention is not limited to the above, but in the second embodiment, the filter coefficient h3
The phase difference corresponding to the delay for each frequency component can be found by Fourier transforming the impulse response of
By adding a predetermined number of delays to 1, it may be possible to specify the positions of the sound sources of a plurality of uncorrelated target sounds a having different frequencies. 21 filter coefficient h3
By setting in advance a value corresponding to a certain delay,
It may be possible to determine the direction of the source of a plurality of target sounds to check the presence or absence of the target sound a.Furthermore, in the above example, two sound receivers are provided, but three or more sound receivers may be used. Without departing from the gist, for example, a module having a sound receiver and an adaptive filter may be used, and these may be arranged two-dimensionally or three-dimensionally to more accurately identify the direction of arrival of the target sound a. It can be implemented with appropriate changes within the scope.

[0024]

[Effects of the Invention] As is clear from the above description, the present invention has a configuration in which two modules each having an adaptive filter are arranged in series, and a mixed sound signal and a reference signal are input and each adaptive filter performs adaptive identification. Accordingly, it is possible to confirm that a signal correlated with the reference signal exists in the mixed sound signal and to extract a target signal correlated with the reference signal from the mixed sound signal, reliably and easily.

[Brief explanation of drawings]

FIG. 1 is a schematic configuration diagram shown to explain the principle of the present invention.

FIG. 2 is a schematic configuration diagram showing a first embodiment of the present invention.

FIG. 3 is a schematic configuration diagram showing a second embodiment of the present invention.

FIG. 4 is a schematic configuration diagram showing a conventional example.

[Explanation of symbols]

11...First module 12...Second module 13, 15...Adaptive filter 17...Reference signal input section 18...Sound receiver 20...Output module a...Target sound a0...Target sound signal b...Non-target sound b0...non-target sound signal u0...Reference signal v…Mixed sound v0...Mixed sound signal

Claims (5)

[Claims] 1. A sound receiver that collects a mixed sound containing a mixture of target sounds and non-target sounds, a reference signal input section into which a reference signal is input, a reference signal from the reference signal input section and the sound receiver. a first module that receives a mixed sound signal from the device, performs adaptive identification using an adaptive filter, estimates and eliminates a target signal correlated with the reference signal in the mixed sound signal, and outputs a non-target sound signal; Then, the output signal of this first module and the mixed sound signal from the sound receiver are inputted and adaptively identified by an adaptive filter, and the non-target signal in the mixed sound signal is estimated and eliminated to produce the target sound. A sound detection device comprising: a second module that outputs a signal. 2. The adaptive filter of the first module is:
2. The sound detection device according to claim 1, wherein the sound detection device converges earlier than the adaptive filter of the second module. 3. A plurality of spaced apart sound receivers that collect a mixture of target sounds and non-target sounds, a reference signal input section into which a reference signal is input, and a reference signal input section corresponding to each of the sound receivers. A mixed sound signal from the sound receiver and a reference signal from the reference signal input section are inputted and adaptively identified by an adaptive filter to generate a target signal having a correlation with the reference signal in the mixed sound signal. a first module that estimates and eliminates the noise signal and outputs a non-target sound signal; and a first module that is provided corresponding to each of the sound receivers and that outputs a mixed sound signal from the sound receiver and an output signal of the first module a plurality of second modules that adaptively identify the input signal using an adaptive filter, estimate and eliminate the non-target signal in the mixed sound signal, and output a target sound signal;
A sound detection device comprising: a third module that performs adaptive identification by changing filter coefficients of an adaptive filter such that a difference between output signals input from the modules is minimized. 4. The adaptive filter of the first module converges before the adaptive filter of the second module, and
4. The sound detection device according to claim 3, wherein the adaptive filter of the third module converges earlier than the adaptive filter of the third module. 5. The sound detection device according to claim 3, wherein the third module is configured to include a filter having a preset filter coefficient.