JPH05259928A - Method and device for canceling adaptive control noise - Google Patents

Abstract

PURPOSE:To satisfactorily exclude the background noises, etc., out of the input acoustic signal. CONSTITUTION:A filter action mode control part 22 actuates a voice source path estimating adaptive filter 16 to estimate a transmission function of the path set between a voice source and the precedent stage of the filter 16 in a period of the desired voices only. Then in a period of the noises only, a noise source path estimating adaptive filter 15 is actuated to estimate the transmission function of the path set between a noise source and the filter 15. Finally the noise alpha.Xb(k) is extracted with use of the transmission functions of both filters 15 and 16 and then subtracted from a main voice signal Xp(k). Thus the noise cancel output is obtained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive control noise canceller device and an adaptive control noise canceling method applicable to, for example, a hands-free automobile telephone.

[0002]

2. Description of the Related Art In recent years, a method for removing noise signal components (for example, engine noise, air conditioner sound, and other background noise) from a voice signal captured by a microphone in a hands-free car telephone or the like has been studied. .. For example, adaptive noise cancellers for such purposes are being developed. For such a technique, for example, Reference 1: May 1985, published by The Institute of Electronics, Information and Communication Engineers, IEICE (A), J69-A, 5, pp 584-591, “Noise Canceller Algorithm Including Preprocessing and Its Performance Evaluation” ], Etc.

Further, techniques such as a sound collecting device for widening the noise removal band of the adaptive control noise canceller are being developed. Regarding this technique, for example, Reference 2:
It is shown in the specification, drawings, etc. of Japanese Patent Application No. 4-4624.

An outline of the operation of the adaptive noise canceller shown in the above document 1 will be described.

According to this document 1, pre-processing and post-processing are performed in different periods, and this pre-processing is performed on the desired signal s.
A known period (k = 0, 1, 2,
..., T) adaptive digital filter 1 (ADF1)
Then, the parameter (transfer function) K (z) of the noise source path (path) from the noise source is estimated. Next, the post-processing is performed by another adaptive digital filter (ADF) in the period other than the above.
According to 2), the noise (n (k)) is calculated by estimating the parameter (transfer function) R (z) of the signal source path from the desired signal source to the direct stage of the ADF 2. The output of this post-processing is configured to obtain a desired signal Xo (k) by subtracting it from the main signal after simple signal processing.

[0006]

However, even the noise removing method according to the above-mentioned document 1 has a problem that it is not possible to obtain a voice signal from which noise superimposed on an input voice signal is sufficiently removed.

This is because, in the post-processing, when the filter coefficient of the ADF2 is updated to obtain the transfer function R (z) of the signal source path, noise n other than the desired signal s (k) is updated.
Since (k) is included, an accurate transfer function R (z) of the signal source path cannot be obtained, and therefore accurate noise calculation cannot be performed, and even if noise is finally subtracted from the main acoustic signal. It is considered that this is because noise cannot be removed sufficiently.

The present invention has been made in view of the above problems, and an object of the present invention is to provide an adaptive control noise canceller device and an adaptive control noise canceler capable of sufficiently removing background noise and the like from an input acoustic signal as compared with the prior art. It is to provide a cancellation method.

[0009]

In order to achieve the above object, the adaptive control noise canceller device according to the first aspect of the present invention is realized by including the following characteristic means.

That is, the main sound capturing means for capturing the sound and outputting the main sound signal, the sub-sound capturing means for capturing the sound and outputting the sub-acoustic signal, and the above-described period during the period when only the desired sound exists. There is only noise and an acoustic source path parameter estimating means for estimating a parameter of a path from a desired acoustic source to the device using the main acoustic signal and the sub acoustic signal, using a first adaptive filter. During the period, a noise source path parameter estimating means for estimating a parameter of a path from a noise source to the device using a second adaptive filter by using the main acoustic signal and the sub acoustic signal, and a desired acoustic And the noise source path parameter and the noise source path parameter are used to perform an operation on the main acoustic signal and the sub-acoustic signal to extract noise and extract the noise. A noise removal operation means for outputting a noise removing signal obtained by removing from the main acoustic signal,
One of the main acoustic source path parameter estimating means for performing the period in which only the desired sound exists or the noise source path parameter estimating means for performing in the period in which only noise exists is executed first, and then the main acoustic source The path parameter estimating means or the other of the noise source path parameter estimating means is executed, and the execution control means for executing execution control is executed after the execution so that the noise removal calculating means is executed in a period in which desired sound and noise exist. It is characterized by being provided.

The adaptive control noise canceling method of the second invention is realized by the following characteristic method in order to achieve the above-mentioned object.

That is, the main sound capturing means for capturing sound and outputting the main sound signal and the sub sound capturing means for capturing sound and outputting the sub sound signal are provided, and only the desired sound exists. One of the acoustic source path parameter estimation processing of the acoustic transmission path from the desired acoustic source during the period or the noise source path parameter estimation processing of the noise transmission path from the noise source during the period when only noise exists is executed first, and then To perform the other of the acoustic source path parameter estimation process, or the noise source path parameter estimation process, during the period when there is a desired sound and noise after this execution, the acoustic source path parameter,
The noise source path parameter is used to perform an operation on the main acoustic signal and the sub acoustic signal to extract noise, and the noise is removed from the main acoustic signal to output a noise-removed signal. It is characterized by doing.

[0013]

According to the first aspect of the present invention, the execution control is performed so that the noise source path parameter is estimated and the acoustic source path parameter is estimated individually, and finally the noise removal calculation means is performed. Since it is controlled by the means, the noise component can be accurately extracted without being affected by the desired acoustic signal as compared with the conventional one, and thus the desired acoustic signal from which the noise is sufficiently removed by the noise removal calculating means. Can be obtained.

According to the second aspect of the invention, the acoustic source path parameter estimating process and the noise source path parameter estimating process are separately performed, and then the acoustic source path parameter estimating process is performed in a period in which desired acoustic and noise are present. Source path parameter and the noise source path parameter, noise is extracted by performing an operation on the main acoustic signal and the sub-acoustic signal, so that the influence of a predetermined acoustic signal is reduced compared to the conventional case. It is possible to accurately extract the noise component without receiving the noise component. Therefore, by removing this noise from the main acoustic signal, it is possible to output an acoustic signal from which the noise component is sufficiently removed.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment of an adaptive control noise canceller device according to the present invention will be described with reference to the drawings.

An object of this embodiment is to realize an adaptive control noise canceller device capable of sufficiently removing background noise and the like from an input acoustic signal as compared with the conventional one.

To this end, the desired speech signal s free of noise (such as the engine sound of a motor vehicle) precedes a known period in which the desired speech signal s (k) is not present.
A period of only (k) (period in which a speaker is generated in a car) is provided, and an adaptive filter is operated in this period to estimate a parameter (transfer function) R (z) of a path from a desired audio signal source. Controlled.

FIG. 1 is a functional block diagram of an adaptive control noise canceller device according to this embodiment.

In FIG. 1, the adaptive control noise canceller device includes a main microphone 10, a sub microphone 11, A / D converters 12 and 13, and an adder 1.
4, a noise source path estimation adaptive filter 15, a speech source path estimation adaptive filter 16, an adder 17, and an adder 1
8, a filter 19, a D / A converter 20, and a filter operation mode controller 22.

The main microphone 10 mainly obtains a desired audio signal s (k) from a desired audio source and A / D converter 1
Supply to 2. The sub microphone 11 mainly obtains the noise n (k) from the noise source and supplies it to the A / D converter 13. A
The / D converter 12 converts the supplied audio signal into a digital signal and outputs the output Xp (k) from the adder 14 and the adder 18.
Supply to. The A / D converter 13 converts the supplied noise into a digital signal and supplies the output Xr (k) to the noise source path estimation adaptive filter 15 and the adder 17.

The adder 14 adds the output Xp (k) and the output of the noise source path estimation adaptive filter 15 and adds the output Xa (k) to the noise source path estimation adaptive filter 15.
The audio source path estimation adaptive filter 16 is supplied. The adder 17 adds using the output Xr (k) and the output Xm (k) of the audio source path estimation adaptive filter 16 to output Xb (k).
(K) is supplied to the filter 19 and the signal source path estimation adaptive filter 16. The filter 19 outputs the coefficient α to the output Xb
The signal α · Xb (k) multiplied by (k) is taken out and supplied to the adder 18. The coefficient α is based on the noise source path estimation adaptive filter 15.

The filter operation mode control unit 22 controls the noise source path estimation adaptive filter 15 and the speech source path estimation adaptive filter 16 to adjust the timing of updating the filter coefficient and the timing of stopping the updating. Supply a signal. This control is controlled by, for example, a control signal in the automobile or a switch operation.

The noise source path estimation adaptive filter 15 is, for example, an FIR (Finite Impulse Resp).
onse) type digital filter or the like, and optimally updates the filter coefficient using the output Xa (k) of the adder 14 and the output Xr (k) of the A / D converter 13 to generate a noise source. The transfer function K (z) of the paths from to the main microphone 10 and the sub microphone 11 is set.

The adaptive filter 16 for estimating the sound source path can also be realized by, for example, an FIR digital filter, and the output Xb (k) of the adder 17 and the adder 14 are used.
Of the output source Xa (k) and the filter coefficient is optimally updated so that the audio source path estimation adaptive filter 16
The transfer function R (z) of the path up to the preceding stage is set.

The adder 18 subtracts the output Xp (k) from the filter output α · Xb (k) and outputs the output Dp.
The signal obtained by the D / A conversion is supplied to the / A converter 20 and can be obtained as the noise cancellation output Xo (k).

FIG. 2 is a processing flowchart of the noise canceling method by the adaptive control noise canceller device according to this embodiment.

In FIG. 2, a desired voice signal without noise is provided as an input signal X (k) as a desired input signal X (k) in a period T (period in which a speaker is generated in a microphone in an automobile without starting an engine). In order to capture only the desired audio signal s (k) from the audio source, the main microphone 10
And an audio signal is captured by the sub microphone 11, and the output Xa (K) of the adder 14 and the output Xb (k) of the adder 17 are captured.
And are supplied to the audio source path estimation adaptive filter 16 (S31). Next, the filter operation mode control unit 22 supplies the update stop control signal of the filter coefficient to the noise source path estimation adaptive filter 15 to stop the update of the filter coefficient, while the audio source path estimation adaptive filter 16 is updated. A control signal is supplied to optimally update the filter coefficient (S
32). Then, the filter coefficient is updated for a predetermined time T (S33), and the optimum transfer function R of the path from the audio source is obtained.
Set (z).

Next, it is judged whether or not the processing is to be ended (S3
4) If it is not finished, the input is made in order to estimate the transfer function of the path from the noise source by the noise source path estimation adaptive filter 15 in a period in which only the noise source exists next (period in which the engine is started). As the signal X (k), the noise n (k) from the noise source is taken in by the main microphone 10 and the sub microphone 11, and the output Xa (k) of the adder 14 and A /
The output Xr (k) of the D converter 13 is supplied to the noise source path estimation adaptive filter 15 (S35). Next, the filter operation mode control unit 22 determines the audio source path estimation adaptive filter 1
The filter coefficient update stop control signal is supplied to 6 to stop the filter coefficient update, and the noise source path estimation adaptive filter 15 is supplied with the update control signal to update the filter coefficient optimally (S36).

Next, as the input signal X (k), a signal is taken in during a period in which a desired voice signal s (k) and noise n (k) exist (a period in which a speaker speaks while the engine is started). (S37), to obtain the noise cancel output,
The filter operation mode control unit 22 outputs the update stop control signal to the noise source path estimation adaptive filter 15 and the voice source path estimation adaptive filter 1 in order to stop the update of the filter coefficient.
6 to stop the update, and in this state, the noise source path estimation adaptive filter 15 and the signal source path estimation adaptive filter 16 are operated with the optimum filter coefficient to obtain the noise cancellation output Xo (k) (S38). ).

FIG. 3 is a characteristic relationship diagram between the input voice signal (main microphone input) and the noise cancel output in the conventional state for explaining the effect of the adaptive control noise canceller device according to this embodiment.

In FIG. 3, it is shown that a noise component is superimposed on the noise cancellation output for the input audio signal obtained by the noise cancellation according to the conventional method.

FIG. 4 is a characteristic relationship diagram between the input voice signal and the noise cancel output for explaining the effect of the adaptive control noise canceller device according to the embodiment.

In FIG. 4, it is shown that the noise component is hardly superimposed on the noise cancel output as compared with FIG.

According to the above embodiment, the transfer function R (z) of the path from the sound source is first estimated, and then the transfer function K (z) of the path from the noise source is estimated. , And then the noise is extracted from the transfer functions R (z) and K (z) estimated above at the time when the desired signal s (k) and the noise n (k) are taken in, and this noise is extracted as the main speech signal. Xp
Since the noise cancellation output Xo (k) is obtained by subtracting from (k), the desired speech signal s (k) and noise n
Accurate noise extraction can be performed without estimating noise while estimating the transfer function of the path from the voice source at the time when (k) is present, and this noise can be accurately removed from the main voice signal.

In FIG. 1 of the above embodiment, the noise source path estimation adaptive filter 15 and the speech source path estimation adaptive filter 16 are realized by FIR type digital filters, but the present invention is not limited to this. For example, IIR (Info
It can also be realized by a digital filter of the unit Impulse Response type.

In FIG. 1 of the above embodiment, two types of microphones are used, but the present invention is not limited to this. For example, it can be realized by three or more types of microphones. For example, when the direction of the noise source is two or more directions, two or more types of sub microphones are provided corresponding to these directions, one main microphone is provided, and the signals captured by the sub microphones are respectively described above. Figure 1
Processing is performed by the configuration of the adaptive filter 16 for estimating the speech source path and the adaptive filter 15 for estimating the noise source path as described above, and finally one adder 18 receives the signals Xp (k) of two systems and the multiplication output α. It can also be applied to supply and subtract Xb (k) to obtain one noise cancel output Xo (k).

In FIG. 2 of the above embodiment, first, the transfer function of the sound source path is estimated by the sound source path estimation adaptive filter 16 (S31 to S33), and then the transfer function of the noise source path. Was operated to be estimated by the noise source path estimation adaptive filter 15 (S35 to S36), but the order is not limited to this. For example, after the transfer function of the noise source path is estimated by the adaptive filter 15 for estimating the noise source path, the transfer function of the voice source path is estimated by the adaptive filter 16 for estimating the signal source path.
You may make it operate | move so that it may estimate.

In the above embodiment, the adaptive control noise canceller for the automobile telephone is explained, but the invention is not limited to this. For example, it can be applied to hearing aids, voice recognition devices, control of precision electronic devices, and the like.

[0039]

As described above, according to the first and second aspects of the invention, the noise source path parameter is estimated and the acoustic source is equipped with the main acoustic capturing means and the sub acoustic capturing means. After performing the estimation of the path parameters individually, finally, during the period when the predetermined sound and noise exist, using the acoustic source path parameter and the noise source path parameter,
Since noise is extracted by performing an operation on the main acoustic signal and the sub acoustic signal, it is possible to accurately extract a noise component without being affected by a predetermined acoustic signal as compared with the conventional art. By removing this noise from the main acoustic signal, it is possible to output an acoustic signal from which noise components have been sufficiently removed.

[Brief description of drawings]

FIG. 1 is a functional block diagram of an adaptive control noise canceller device according to an embodiment of the present invention.

FIG. 2 is a processing flowchart of a noise canceling method by the adaptive control noise canceller device according to the embodiment of the present invention.

FIG. 3 is a characteristic relationship diagram between an input voice signal and a noise cancel output in a conventional state for explaining the effect of the adaptive control noise canceller device according to the embodiment of the present invention.

FIG. 4 is a characteristic relationship diagram between an input audio signal and a noise cancel output for explaining the effect of the adaptive control noise canceller device according to the embodiment of the present invention.

[Explanation of symbols]

10 ... Main microphone, 11 ... Sub microphone, 1
2 to 13 ... A / D converter, 14, 17 ... Adder, 15 ...
Adaptive filter for noise source path estimation, 16 ... Adaptive filter for signal source path estimation, 18 ... Adder, 19 ... Filter, 20
... D / A converter, 22 ... Filter operation mode control unit.

 ─────────────────────────────────────────────────── ─── Continued Front Page (72) Inventor Shinsuke Takada 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd.

Claims (2)

[Claims] 1. A main sound capturing unit that captures sound and outputs a main sound signal; a sub sound capturing unit that captures sound and outputs a sub sound signal; An acoustic source path parameter estimating means for estimating a parameter of a path from a desired acoustic source to the device by using the main acoustic signal and the sub-acoustic signal and using a first adaptive filter, and only noise exists. A noise source path parameter estimating means for estimating a parameter of a path from a noise source to the device by using the second adaptive filter by using the main acoustic signal and the sub acoustic signal during the period; During the period when noise is present, the acoustic source path parameter and the noise source path parameter are used to perform an operation on the main acoustic signal and the sub acoustic signal to extract noise, A noise removal calculation means for outputting a noise removal signal obtained by removing from the main acoustic signal, and the above-mentioned acoustic source path parameter estimation means for performing the period when only the desired sound exists, or for the period when only noise exists. One of the noise source path parameter estimation means is executed first, and then the other of the acoustic source path parameter estimation means or the noise source path parameter estimation means is executed, and after this execution, desired sound and noise exist. An adaptive control noise canceller device, comprising: an execution control unit for executing execution control for causing the noise removal calculation unit to be executed during a period. 2. A main sound capturing means for capturing sound and outputting a main sound signal, and a sub sound capturing means for capturing sound and outputting a sub sound signal, wherein only the desired sound exists. Acoustic source path parameter estimation processing of the acoustic transfer path from the desired acoustic source during the period, or
One of the noise source path parameter estimation processing of the noise transfer path from the noise source is executed first in the period when only the noise exists, and then the other of the acoustic source path parameter estimation processing or the noise source path parameter estimation processing is performed. After the execution, during the period when the desired sound and noise are present after this execution, using the acoustic source path parameter and the noise source path parameter, the main acoustic signal and the auxiliary acoustic signal are calculated. An adaptive control noise canceling method, which comprises: performing noise extraction, removing the noise from the main acoustic signal, and outputting the noise-free signal.