JPH04227338A - Voice signal processing unit - Google Patents

Abstract

PURPOSE:To realize the voice signal processing unit improving the S/N of a voice signal including noise. CONSTITUTION:A voice input including noise is divided into respective bands by a band split means 11. Then a voice band detection means 12 discriminates to which band the voice is included. Then a band selection emphasis control means 13 selects a band (in which a voice signal is included) to be emphasized and a voice standby selection emphasis means 14 emphasizes the signal for each band at an optimum ratio. Then a band synthesis means 15 synthesizes the bands and outputs the result finally.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal processing device that can be used for audio processing and the like.

0002

2. Description of the Related Art FIG. 25 is a block diagram of a conventional signal processing device. 1 is a filter control unit to which a signal mixed with noise is input and detects the signal or noise; 2 is a BPF group having a large number of bandpass filters; and 3 is an adder. That is, the filter control section 1 controls the filter coefficients of the BPF group 2 according to the noise or signal of the input signal, and the BPF group 2 is a band-pass filter group that divides the input signal into appropriate bands, It is configured such that its pass band characteristics are determined by a control signal from the filter control section 1. The operation of the conventional signal processing device configured as described above will be described below.

[0003] An input signal in which noise is superimposed on speech is supplied to a filter control section 1 . The filter control section 1 obtains noise components from the supplied signal corresponding to each band of the BPF group 2, and supplies the BPF group 2 with filter coefficients that do not allow the noise components to pass through the BPF group 2.

The BPF group 2 divides the input signal into appropriate bands, passes the input signal appropriately according to the filter coefficients inputted from the filter control section 1 for each band, and passes the input signal appropriately to the adder 3.
supply to. The adder 3 mixes the signals divided into appropriate bands by the BPF group 2 to obtain an output.

[0005] By doing this, the input signal becomes B
PF group 2 lowers the pass level of the band containing noise. As a result, a signal with attenuated noise components is obtained.

[0006]

[Problems to be Solved by the Invention] However, such noise attenuation methods still have the problem that noise removal is not sufficient.

[0007] The present invention solves the problems of the conventional audio signal processing apparatus, and aims to provide an audio signal processing apparatus that can effectively suppress noise.

Furthermore, in conventional noise suppression methods, the noise portion and the voice portion are simply distinguished in terms of time, and the entire portion is attenuated or amplified. This means that the S/N ratio in each part is not particularly improved. Therefore, an object of the present invention is to provide an audio signal processing device that improves the S/N ratio in consideration of the problems of the conventional noise suppression method.

[0009]

[Means for Solving the Problems] In order to achieve the above object, the present invention, as a first aspect, includes a band dividing means for inputting a mixed signal containing noise and dividing the band into bands; A voice band detecting means for detecting the voice band portion of the signal for each channel, and a noise-mixed signal that has been band-divided based on the voice band information detected by the voice band detecting means.
The present invention is characterized by comprising a voice band selection and emphasis means that emphasizes the voice signal band relative to the noise signal band, and a band synthesis means that performs band synthesis on the signal emphasized by the voice band selection and emphasis means. This is an audio signal processing device.

[0010] In addition, as a second aspect of the present invention, a band dividing means inputs a mixed signal containing noise and divides the band, and a voice part of the band divided signal is determined by the band dividing means. a noise prediction means for predicting noise in the voice part using the discriminated voice part information; a canceling means for subtracting the predicted noise value from the band-divided signal; audio band detection means for detecting an audio band portion of each band-divided signal for each channel; and a signal from which noise has been canceled by the canceling means based on the audio band information detected by the audio band detection means. and a band synthesis means for band-synthesizing the signals emphasized by the voice band selection and emphasis means. This is an audio signal processing device characterized by:

Further, as a third aspect of the present invention, there is provided a band dividing means for dividing the frequency band of an inputted audio signal including noise, and a band dividing output of the band dividing means is input, and the noise component is predicted. a pitch frequency detection means for detecting the pitch frequency of the input signal including the noise; a cancellation coefficient setting means for setting a cancellation coefficient according to the pitch frequency output of the pitch frequency detection means; and the noise prediction means. A noise prediction output of the means, a band division output of the band division means, and a cancellation coefficient signal set by the cancellation coefficient setting means are input, and from the band division output, a noise component is canceled in consideration of the cancellation rate. means, a voice band detecting means for detecting a voice band portion of the input signal using the pitch frequency detected by the pitch frequency detecting means, and a control signal for emphasizing the voice band detected by the voice band detecting means. and a band selection emphasis control means for outputting an audio signal that, in accordance with a control signal of the band selection emphasis control means, emphasizes the audio signal band relative to the noise signal band for the signal whose noise has been canceled by the canceling means. Regarding the band selective emphasis means and the signal emphasized by the audio band selective emphasis means,
The present invention is an audio signal processing device characterized by comprising a band synthesis means for performing band synthesis.

Furthermore, the present invention in a fourth aspect includes a band dividing means for dividing the frequency band of an inputted audio signal including noise, and a band dividing output of the band dividing means is input, and the noise component is predicted. a pitch frequency detection means for detecting the pitch frequency of the input signal including the noise; a cancellation coefficient setting means for setting a cancellation coefficient according to the pitch frequency output of the pitch frequency detection means; and the noise prediction means. A noise prediction output of the means, a band division output of the band division means, and a cancellation coefficient signal set by the cancellation coefficient setting means are input, and from the band division output, a noise component is canceled in consideration of the cancellation rate. a voice band detecting means for detecting a voice band portion of the input signal using the pitch frequency detected by the pitch frequency detecting means; and a voice band detected by the voice band detecting means. Noise is canceled by a noise band calculating means for calculating a noise band based on the information, a band selection attenuation control means for outputting a control signal for attenuating the noise band calculated by the noise band calculating means, and the canceling means. a noise band selection attenuation means which inputs the signal, selects a noise band and attenuates only that noise band in accordance with a control signal of the band selection attenuation control means, and the signal attenuated by the noise band selection attenuation means, The present invention is an audio signal processing device characterized by comprising a band synthesis means for performing band synthesis.

[0013]

[Operation] In the device of the first aspect of the present invention having the above configuration, the band dividing means inputs a mixed signal containing noise and divides the band, and the audio band detecting means inputs the mixed signal containing noise. The audio band selection and emphasis means detects the audio band portion of the signal for each channel, and based on the audio band information detected by the audio band detection means, the audio band portion of the signal mixed with noise that has been divided into bands is The signal band is emphasized relative to the noise signal band, and the band synthesis means performs band synthesis on the signal emphasized by the voice band selection and emphasis means. Relatively emphasizing the voice signal band means increasing the signal level of the voice signal band or decreasing the signal level of the noise signal band.

Further, in the second aspect of the apparatus of the present invention, the band dividing means inputs a mixed signal including noise, divides the signal into bands, and the voice discriminating means uses the band dividing means to divide the mixed signal into the band divided signals. Then, the voice part is determined, the noise prediction means uses the determined voice part information to predict the noise of the voice part, and the canceling means calculates the predicted noise value from the band-divided signal. On the other hand, the audio band detecting means detects the audio band part of the band-divided signal for each channel, and the audio band detecting means detects the audio band part of the band-divided signal for each channel. Based on the audio band information detected by the audio band detecting means, the audio band selection and emphasis means calculates the audio signal band relative to the noise signal band for the signal whose noise has been canceled by the canceling means. The band synthesis means performs band synthesis on the signal emphasized by the audio band selective emphasis means.

Further, in the third aspect of the apparatus of the present invention, the band dividing means divides the frequency band of the input audio signal containing noise, and the noise predicting means inputs the band divided output, and predicting the noise component and detecting the pitch frequency of the input signal including the noise with a pitch frequency detection means;
The cancellation coefficient setting means sets a cancellation coefficient according to the pitch frequency output, the cancellation means inputs the noise prediction output, the band division output, and the cancellation coefficient signal, and cancels it from the band division output. the noise component in consideration of the frequency, the voice band detecting means detects the voice band portion of the input signal using the pitch frequency, and the band selection emphasis control means generates a control signal for emphasizing the voice band. According to the control signal, the audio band selection and emphasis means outputs a signal in which the noise has been canceled,
Emphasizes the voice signal band relative to the noise signal band,
The band synthesis means adds to the emphasized signal,
Perform band synthesis.

Further, in the fourth aspect of the apparatus of the present invention, the band dividing means divides the frequency band of the input audio signal containing noise, and the noise predicting means inputs the band divided output, and predicting the noise component and detecting the pitch frequency of the input signal including the noise with a pitch frequency detection means;
The cancellation coefficient setting means sets a cancellation coefficient according to the pitch frequency output, the cancellation means inputs the noise prediction output, the band division output, and the cancellation coefficient signal, and cancels it from the band division output. canceling the noise component in consideration of the frequency, the voice band detection means detects the voice band part of the input signal using the pitch frequency, and the noise band calculation means calculates the noise band from the voice band part. , the band selection emphasis control means outputs a control signal for attenuating the noise band, and the noise band selection attenuation means adjusts the noise signal band to the voice signal band for the noise-cancelled signal according to the control signal. is relatively attenuated, and by means of band synthesis, the signal is
Perform band synthesis.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Examples of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of a first embodiment of the audio signal processing apparatus of the present invention. In FIG. 1, the band dividing means 11 inputs a signal containing a mixture of voice and noise, and
After the conversion, it is a means of performing Fourier transformation.

The audio band detecting means 12 is a means for inputting the mixed signal containing noise, which has been Fourier transformed and outputted by the band dividing means 11, and detecting the band of the audio signal portion in the mixed signal. For example, the cepstrum (
CEPSTRUM) analysis is used to detect the band in which the signal is present. The relationship between the voice band and the noise band based on frequency is normally as shown in FIG. Here, S is the audio signal band,
N is the noise band. This audio band S is detected.

The band selection emphasis control means 13 is a means for outputting a control signal for emphasizing a voice band based on the voice band information detected by the voice band detection means 12.

The voice band selection and emphasis means 14 inputs the noise-containing signal band-divided by the band division means 11, selects a voice band according to the control signal of the band selection and emphasis control means 13, and emphasizes only that band. It is a means to do so.

The band synthesizing means 15 is a means for band synthesizing the signal emphasized by the audio band selection emphasizing means 4.

Next, the operation of the first embodiment will be explained. Band division means 11 performs band division on the noise-containing audio signal. The audio band detection means 12 detects the audio band of the band-divided signal. Band selection emphasis control means 13 generates a control signal based on the audio band information. The audio band selection emphasizing means 14 emphasizes the audio level of the audio band of the band-divided audio-mixed signal in accordance with the control signal from the band selection emphasizing control means 13. The band synthesis means 15 performs band synthesis on this voice level-enhanced voice signal mixed with noise and outputs the result.

FIG. 2 is a block diagram showing a further embodiment of the audio band detecting means 12 of the first embodiment shown in FIG. That is, the voice band detection means 12 includes a cepstrum analysis means 21, a peak detection means 22, and a voice band detection means 23. The cepstrum analysis means 21 is a means for performing cepstrum analysis on the signal that has been Fourier transformed by the band division means 11. The cepstrum is the inverse Fourier transform of the logarithm of the short-time amplitude spectrum of a waveform, as shown in FIG. FIG. 20(A) is a short-time spectrum,
Figure (B) is the cepstrum. The peak detection means 22 is a means for detecting the peak (pitch) of the cepstrum determined by the cepstrum analysis means 21 and distinguishing between a signal and noise. The portion where the peak exists is determined to be the audio signal portion. The peak can be detected, for example, by determining a threshold value of a predetermined size in advance and comparing it with the threshold value. Furthermore, the audio band detection circuit 23 has a peak detection means 22.
This is a means for determining the quefrency value of the detected peak from FIG. 20(B). This allows the audio band to be detected. Since the other means are the same as those in the embodiment shown in FIG. 1, their explanation will be omitted.

FIG. 3 is a block diagram showing another embodiment in which the audio band detecting means 12 of the embodiment of FIG. 1 is further embodied. That is, the voice band detection means 12 of this embodiment includes a formant analysis means 24 in addition to a cepstrum analysis means 21, a peak detection means 22, and a voice band detection circuit 23. This formant analysis means 24 is a means for analyzing formants in the cepstrum analysis results obtained by the cepstrum analysis means 21 (Fig. 20(
See B). The voice band detection circuit 23 detects a voice band using both the peak information detected by the peak detection means 22 and the formant information analyzed by the formant analysis means 24. In this embodiment, since the voice band is detected using formant information in addition to the peak information, it is possible to detect the voice band more precisely. The other means are the same as those in the embodiment shown in FIG. 2, so their explanation will be omitted.

FIG. 4 is a block diagram showing an embodiment of the present invention, which is an apparatus for calculating a noise band and attenuating the noise level of the noise band.

The band division means 11, cepstrum analysis means 21, peak detection means 22, and voice band detection circuit 23 in this embodiment are the same as those in the embodiment of FIG.
The explanation will be omitted.

The output of the voice band detection circuit 23 is input to the noise band calculation means 6. The noise band calculating means 16 is a means for calculating a noise band based on the detected voice band information, and is a means for determining, for example, a band other than the voice band as a noise band. The band selection attenuation control means 17 is a means for outputting an attenuation control signal based on the noise band information calculated by the noise band calculation means 16. The noise band selection attenuation means 18 has its band selection attenuation control means 1
This is a means for attenuating the signal level of the noise band in the band-divided signal sent from the band dividing means 11 according to the control signal from the band dividing means 7. As a result, the audio band signal is relatively emphasized. The band synthesis means 15 band-synthesizes the signals whose signal levels in the noise band have been attenuated. In this embodiment, since the signal level is attenuated in the noise band, the voice band is relatively emphasized, and the S/N ratio is improved.

FIG. 5 shows the embodiment of FIG. 4 with formant analysis means added. In this example as well, the voice band can be detected more precisely by formant analysis, so
The noise band calculation means 16 detects the noise band more precisely.

FIG. 16 shows a combination of the embodiment shown in FIG. 2 and the embodiment shown in FIG. That is, the band dividing means 1
1, cepstrum analysis means 21, peak detection means 22,
The audio band detection circuit 23 is provided in common. The output of the voice band detection circuit 23 is input to the band selection emphasis control means 3 and the noise band calculation means 16, respectively. The output of the band selection emphasis control means 13 is input to the audio band selection emphasis means 14. Then, the audio band selection and emphasis means 14 amplifies the signal level of only the audio band of the band-divided signal. On the other hand, the noise band calculation means 16
The noise band calculated in is input to the band selection attenuation control means 17, and a control signal is output from there. The noise band selection and attenuation means 18 attenuates the signal level of only the noise band of the output of the voice band selection and emphasis means 14, and then attenuates the signal level of the voice band. Note that this voice band selection emphasis means 4 and noise band selection attenuation means 1
In No. 8, the signal processing may be reversed, that is, the signal level of the noise band may be attenuated first, and then the signal level of the voice band may be amplified. This audio band selection emphasis means 14
and the noise band selection attenuation means 18 constitute an emphasis/attenuation means 19. In this embodiment, since the sound level in the voice band is amplified and the noise level in the noise band is attenuated at the same time, it is possible to further improve the S/N ratio.

FIG. 7 shows an embodiment in which a formant analysis means 24 is added to the embodiment of FIG. 6, the contents of which are as described above, and the other means are the same as those of FIG. 6, so their explanation will be omitted. . By adding formant analysis means, more precise detection of voice bands becomes possible.

In the embodiments described above, the voice band detection means, voice band selection and emphasis means, etc. can be implemented as software using a computer, but it is also possible to use dedicated hardware circuits having the respective functions. is also possible.

[0033] As is clear from the above explanation,
In the above embodiment of the present invention, a voice signal mixed with noise is band-divided, and the voice level of the voice band is emphasized relative to the signal level of the noise band for the band-divided signal. Therefore, the effect is that the S/N ratio is significantly improved. FIG. 8 is a block diagram of another embodiment of the audio signal processing device of the present invention.

In the figure, the band dividing means 11 is a means for inputting a signal containing a mixture of voice and noise, performing A/D conversion, and then Fourier transform.

The audio band detecting means 12 is a means for inputting the mixed signal containing noise, which has been Fourier transformed and outputted by the band dividing means 11, and detecting the band of the audio signal portion in the mixed signal. For example, the cepstrum (
It has a speech analysis means 21 that performs CEPSTRUM analysis, and a speech band detection circuit 23 that detects a speech band using the result of the cepstrum analysis. This audio band and
The relationship with the noise band based on frequency is normally as shown in FIG. Here, S is the audio signal band and N is the noise band. This audio band S
is detected.

The band selection emphasis control means 13 is a means for outputting a control signal for emphasizing a voice band based on the voice band information detected by the voice band detection means 23.

The speech discriminating means 31 is a means for discriminating the speech part of the band-divided speech signal mixed with noise. and a voice discrimination circuit 32 that discriminates voice signals using the voice discrimination circuit 32.

The noise prediction means 33 includes the band division means 1
1, the noise portion is detected from the voice portion discriminated by the voice discriminating means 31, and the noise of the voice portion is predicted based on the noise data of the noise-only portion. The noise prediction means 33 is a means for predicting noise components for each channel based on voice/noise input divided into m channels. For example, as shown in FIG. 22, the x-axis shows the frequency, the y-axis shows the audio level, and the z-axis shows the time, and the data p1 at the frequency f1,
Take p2, . . . , pi and predict the next pj. For example, the average of the noise parts p1 to pi is taken as pj. Alternatively, when the audio signal portion continues, pj is multiplied by an attenuation coefficient.

The canceling means 34 includes the band dividing means 11
m-channel signals are supplied from the noise prediction means 33, and the noise is canceled by subtracting the noise for each channel. As shown in FIG. 23, the cancellation method is based on frequency, and the noise-containing audio signal (A) is Fourier transformed (C
), then subtract the predicted noise spectrum (D) (E)
, which is inversely Fourier transformed to obtain a noise-free audio signal (F).

The audio band selection emphasis means 14 inputs the noise-containing audio signal from which the noise has been removed to some extent by the cancellation means 34, and the audio band selection emphasis control means 1
This is a means for selecting an audio band and emphasizing it according to the control signal No. 3.

The band synthesizing means 15 is a means for band synthesizing the signals emphasized by the audio band selection emphasizing means 14. For example, inverse Fourier transform.

Next, the operation of this embodiment will be explained. Band division means 11 performs band division on the noise-containing audio signal. The audio band detection means 12 detects the audio band of the band-divided signal. Band selection emphasis control means 13 generates a control signal based on the audio band information.

On the other hand, the speech discrimination means 31 predicts the noise in the speech signal part of the band-divided speech signal mixed with noise. The canceling means 34 removes the predicted noise value from the band-divided noisy audio signal. Then, the audio band selection emphasis means 14 relatively emphasizes the audio level of the audio band based on the control signal from the band selection emphasis control means 13 for the signal from which this noise has been removed to some extent.

The band synthesizing means 15 performs band synthesis on this voice level-enhanced and noise-mixed audio signal and outputs the result.

FIG. 9 is a block diagram showing a further embodiment of the speech analysis means 21 of the embodiment shown in FIG. That is, the speech analysis means 21 is the cepstrum analysis means 2.
1 and a peak detection means 22. The cepstrum analysis means 21 uses the band division means 11 to
This is a means of performing cepstral analysis on Fourier-transformed signals. The cepstrum is the inverse Fourier transform of the logarithm of the short-time amplitude spectrum of a waveform, as shown in FIG. FIG. 12(A) is a short-time spectrum, and FIG. 12(B) is its cepstrum. The peak detection means 22 includes the cepstrum analysis means 2.
Cepstrum peak (pitch) determined in 1
This is a means of detecting and distinguishing between voice signals and noise signals. The portion where the peak exists is determined to be the audio signal portion. The peak can be detected, for example, by determining a threshold value of a predetermined size in advance and comparing it with the threshold value. Furthermore, the audio band detection circuit 23
This is a means for obtaining a quefrency value at a peak detected by the peak detecting means 22 from FIG. 20(B). This allows the audio band to be detected. Further, the voice discrimination circuit 32 discriminates the voice signal portion from the peak detected by the peak detection means 22. Since the other means are the same as those in the embodiment shown in FIG. 8, their explanation will be omitted.

FIG. 10 shows that the embodiment shown in FIG. 9 is further provided with formant analysis means 24. This formant analysis means 24 is a means for analyzing formants in the cepstrum analysis results obtained by the cepstrum analysis means 21 (see FIG. 20(B)). The voice band detection circuit 23 detects a voice band using both the peak information detected by the peak detection means 22 and the formant information analyzed by the formant analysis means 24. In this embodiment, since the voice band is detected using formant information in addition to the peak information, it is possible to detect the voice band more precisely. The other means are the same as those in the embodiment shown in FIG.

FIG. 11 is a block diagram showing an embodiment of the present invention, which is an apparatus for calculating a noise band and attenuating the noise level of the noise band.

The band division means 11, cepstrum analysis means 21, peak detection means 22, and voice band detection circuit 23 in this embodiment are the same as those in the embodiment of FIG.
The explanation will be omitted.

The output of the voice band detection circuit 23 is input to the noise band calculation means 16. The noise band calculation means 16 is
It is a means for calculating a noise band based on the detected voice band information, and is a means for determining, for example, a band other than the voice band as a noise band. The band selection attenuation control means 17 is a means for outputting an attenuation control signal based on the noise band information calculated by the noise band calculation means 16. The noise band selection attenuation means 18 is a means for attenuating the signal level of the noise band in the signal sent from the cancellation means 34 in accordance with the control signal from the band selection attenuation control means 17. As a result, the audio band signal is relatively emphasized. The band synthesis means 15 band-synthesizes the signals whose signal levels in the noise band have been attenuated. In this embodiment, since the signal level is attenuated in the noise band, the voice band is relatively emphasized, and the S/
N ratio improves.

FIG. 12 shows the embodiment shown in FIG. 11 in which a formant analysis means 24 is added. In this embodiment as well, since the voice band is detected more precisely by formant analysis, the noise band calculation means 16 detects the noise band more precisely.

FIG. 13 shows a combination of the embodiment shown in FIG. 9 and the embodiment shown in FIG. That is, band division means 11, cepstrum analysis means 21, peak detection means 22
, the voice discrimination circuit 32, and the voice band detection circuit 23 are provided in common. The output of this audio band detection circuit 23 is
The signals are input to the band selection emphasis control means 13 and the noise band calculation means 16, respectively. Band selection emphasis control means 1
The output of No. 3 is input to the audio band selection and emphasis means 14. The audio band selection emphasis means 14 then cancels the cancellation means 3.
The signal level of only the audio band of the signal output from 4 is amplified. On the other hand, the noise band calculated by the noise band calculation means 16 is input to the band selection limit control means 17,
A control signal is output from there. The noise band selection and attenuation means 18 attenuates the signal level of only the noise band of the output of the voice band selection and emphasis means 14. In addition,
The voice band selective emphasis means 14 and the noise band selective attenuation means 18 may perform signal processing in the opposite manner, that is, first attenuate the signal level of the noise band and then amplify the signal level of the voice band. This audio band selection emphasis means 1
4 and the noise band selection attenuation means 18 are the emphasis/attenuation means 35
Configure. In this embodiment, since the sound level in the voice band is amplified and the noise level in the noise band is attenuated at the same time, it is possible to further improve the S/N ratio.

FIG. 14 shows an embodiment in which a predetermined restriction is applied to the function of the band selection emphasis control means 13 of the embodiment shown in FIG. 9 to make the improvement of the S/N ratio reasonable.

That is, the noise power calculation means 37 is a means for calculating the magnitude of noise based on the output of the noise prediction means 33. The audio signal power calculation means 36 is a means for calculating the magnitude of the audio signal emphasized by the audio band selection and emphasis means 14. The S/N calculation means 38 is a means for inputting the audio signal calculated by the audio signal power calculation means 36 and the noise power calculated by the noise power calculation means 37, and calculates the S/N ratio. . Then, the band selection emphasis control means 13 controls the S/N
The ratio calculation means 38 inputs the calculated S/N ratio and outputs a control signal to the audio band selection emphasis means 14 so that the S/N ratio becomes a predetermined target S/N value. This target S/N value is, for example, 1/5. This target S/N value has the meaning of preventing the audio signal from being overemphasized compared to the noise.

FIG. 15 shows the S/N ratio by applying predetermined constraints to the function of the band selection attenuation control means 17 of the embodiment shown in FIG.
This is an example that makes the ratio improvement reasonable.

That is, as described above, the noise power calculation means 37 is a means for calculating the magnitude of noise based on the output of the noise prediction means 33. The audio signal power calculation means 36 is a means for calculating the magnitude of the audio signal that is relatively emphasized due to the noise being attenuated by the noise band selection attenuation means 18. The S/N calculation means 38 is
This means inputs the audio signal calculated by the audio signal power calculation means 36 and the noise power calculated by the noise power calculation means 37, and calculates the S/N ratio. Then, the band selection attenuation control means 17 controls the S/N
The ratio calculation means 38 inputs the calculated S/N ratio and outputs a control signal to the audio band selection emphasis means 14 so that the S/N ratio becomes a predetermined target S/N value.

In the embodiments described above, the voice band detection means, voice band selection and emphasis means, etc. can be realized in software using a computer, but it is also possible to use dedicated hardware circuits having the respective functions. is also possible.

As is clear from the above description of the embodiments, the present invention is applicable to noise-containing audio signals.
Perform band division, cancel the noise predicted for the band-divided signal, and compare the audio level of the voice band relative to the signal level of the noise band for the signal whose noise has been canceled to some extent. S
This has the effect of significantly improving the /N ratio.

FIG. 16 is a block diagram of an audio signal processing device in another embodiment of the present invention. In FIG. 16, reference numeral 11 denotes a band division means for dividing a noise-containing audio signal into frequency bands, as an example of a frequency analysis means for analyzing the frequency of a signal, and 33 denotes the band division means 11.
Noise prediction means for inputting the output of and predicting noise components; 4
1 is a canceling means for removing noise as described later, and 15 is a band synthesizing means for synthesizing voices, which is an example of a signal synthesizing means for synthesizing signals.

To explain in detail, the band dividing means 11 is a means to which a voice/noise input containing noise is supplied, performs band division into m channel frequency bands, and supplies the divided frequency bands to the noise predicting means 33 and the canceling means 41. Noise prediction means 3
3 is means for predicting noise components for each channel based on the voice/noise input divided into m channels and supplying the predicted noise components to the canceling means 41. For example, the noise prediction is as shown in FIG. In other words, the x-axis is the frequency, the y-axis is the audio level, and the z-axis is time.
Data p1, p2, ..., pi at frequency f1
, and predict the future pj. For example, the noise part p
The average of 1 to pi is taken as pj. Alternatively, when the audio signal portion continues, pj is multiplied by an attenuation coefficient. Further, the canceling means 41 is supplied with m-channel signals from the band dividing means 11 and the noise predicting means 33, and performs cancellation by subtracting noise for each channel according to the input of a cancellation coefficient. That is, the predicted noise component is multiplied by a cancellation coefficient to cancel it. Generally, as an example of a cancellation method, cancellation on the time axis involves subtracting a predicted noise waveform (B) from a noise-containing audio signal (A), as shown in FIG. Only the signal is thereby extracted (F).

In this embodiment, as shown in FIG. 23, cancellation is performed on the basis of frequency, and the noisy speech signal (A) is Fourier transformed, (C), and then the predicted noise spectrum (D) is (E) and performs inverse Fourier transform on it to obtain a noise-free audio signal (F).
The pitch frequency detection means 42 is means for detecting the pitch frequency of the input voice/noise input and supplying it to the cancellation coefficient setting means 43. Here, the pitch frequency of the voice is determined by various methods as shown in Table 1, and represents the presence or absence of voice and its characteristics.

It should be noted that instead of the pitch frequency detection means 42, means for detecting other audio portions may be provided.

The cancellation coefficient setting means 43 sets m cancellation coefficients based on the pitch frequency supplied from the pitch frequency detection means 42, and
It is configured to supply

The audio band detecting means 23 is a means for detecting the band of the audio signal portion using the pitch frequency detected by the pitch frequency detecting means 42. For example, the audio band is detected using the cepstral analysis results. The relationship between the voice band and the noise band based on frequency is normally as shown in FIG. 21. Here, S is the audio signal band and N is the noise band. This audio band S is detected.

The band selection emphasis control means 13 is a means for outputting a control signal for emphasizing a voice band based on the voice band information detected by the voice band detection means 23. The audio band selection emphasizing means 14 includes the canceling means 4
1 inputs a noisy audio signal from which noise has been removed to some extent, selects an audio band according to a control signal from the band selection emphasis control means 13, and emphasizes it.

The band synthesizing means 15 is a means for band synthesizing the signals emphasized by the audio band selection emphasizing means 14. For example, an inverse Fourier transformer.

The operation of the signal processing device according to the embodiment of the present invention constructed as described above will be explained below.

The voice/noise input containing noise is band-divided into m-channel signals by the band division means 11, and the noise component is predicted for each channel by the noise prediction means 33. The signal band-divided into m channels by the band division means 11 is subjected to cancellation means 41 in which noise components supplied from the noise prediction means 33 are removed for each channel. The noise removal rate at this time is appropriately set by inputting a cancellation coefficient to improve clarity for each channel. For example, even if there is noise where there is an audio signal, the clarity is improved by reducing the cancellation coefficient so that the noise is not removed much. More specifically, the noise removal rate at this time is determined by the cancellation coefficient setting means 43.
It is set for each channel by the cancellation coefficient supplied by That is, if the predicted noise component is ai, the noise-mixed signal is bi, and the cancellation coefficient is αi, then the output ci of the canceling means 41 is (bi-αi×ai). On the other hand, the cancellation coefficient is determined based on information from the pitch frequency detection means 42. That is, audio/
The pitch frequency detection means 42 receives the noise input and detects the pitch frequency of the voice. Then, the cancellation coefficient setting means 43 sets a cancellation coefficient as shown in FIG. That is, FIG. 24(A) shows the cancellation coefficient in each band. Here, f0-f3 indicates the entire band of voice/noise input, and this f0-f3 is
Divide into channels and set the cancellation coefficient. f
1-f2 particularly indicates a band in which voice is included, and is determined using the pitch frequency. In this way, in the voice band, the cancellation coefficient is made small (close to 0) and noise is not removed as much as possible. This improves clarity. This is because human hearing can hear voices even if there is some noise. In the non-voice bands of f0-f1 and f2-f3, the cancellation coefficient is set to 1 to sufficiently remove noise. The cancellation coefficient in FIG. 24(B) is used when it is certain that there is no voice at all and that it can only be considered as noise, and is set to 1 to sufficiently remove noise. For example, if a vowel does not appear at all in terms of peak frequency, it cannot be determined that it is a voice signal, so it is determined to be noise. It is desirable to be able to switch the cancellation coefficients in FIGS. 24(A) and 24(B) as appropriate.

On the other hand, based on the pitch frequency information detected by the pitch frequency detection means 42, the voice band detection means 23 detects the voice band. Band selection emphasis control means 13 generates a control signal based on the audio band information. Then, the audio band selection emphasis means 14 relatively emphasizes the audio level of the audio band based on the control signal from the band selection emphasis control means 13 for the signal from which noise has been removed to some extent by the cancellation means 41. do.

[0069] The band synthesis means 15 performs band synthesis on this voice level-enhanced voice signal mixed with noise and outputs the result.

FIG. 17 is a block diagram showing another embodiment of the present invention, which differs from the embodiment of FIG. 16 in that a noise band is calculated and the noise level of the noise band is attenuated.

That is, the band division means 11, noise prediction means 33, cancellation means 41, pitch frequency detection means 42, cancellation coefficient setting means 43, and voice band detection means 23 in this embodiment are the same as those in the embodiment shown in FIG. Since they are the same, their explanation will be omitted.

The output of the voice band detection means 23 is input to the noise band calculation means 16. The noise band calculation means 16 is
It is a means for calculating a noise band based on the detected voice band information, and is a means for determining, for example, a band other than the voice band as a noise band. The band selection attenuation control means 17 is a means for outputting an attenuation control signal based on the noise band information calculated by the noise band calculation means 16. The noise band selection attenuation means 18 is a means for attenuating the signal level of the noise band in the signal sent from the cancellation means 41 in accordance with the control signal from the band selection attenuation control means 17. As a result, the audio band signal is relatively emphasized.

In this embodiment, since the signal level is attenuated in the noise band, the voice band is relatively emphasized, and the S/N ratio is improved.

FIG. 18 shows the S/N ratio by applying predetermined constraints to the function of the band selection emphasis control means 13 of the embodiment shown in FIG.
This is an example that makes the ratio improvement reasonable.

That is, the noise power calculation means 37 is a means for calculating the magnitude of noise based on the output of the noise prediction means 33. The audio signal power calculation means 36 is a means for calculating the magnitude of the audio signal emphasized by the audio band selection and emphasis means 14. The S/N calculating means 38 is a means for inputting the audio signal calculated by the audio signal power calculating means 36 and the noise power calculated by the noise power calculating means 37, and calculates the S/N ratio. Then, the band selection emphasis control means 13 inputs the S/N ratio calculated by the S/N ratio calculation means 38,
A control signal such that the S/N ratio becomes a predetermined target S/N value is output to the audio band selection and emphasis means 14. This target S/N value is, for example, 1/5. This target S/N value has the meaning of preventing the audio signal from being overemphasized compared to the noise.

FIG. 19 shows the S/N ratio by applying predetermined constraints to the function of the band selective attenuation control means 17 of the embodiment shown in FIG.
This is an example that makes the ratio improvement reasonable.

That is, as described above, the noise signal power calculation means 37 is a means for calculating the magnitude of noise based on the output of the noise prediction means 33. The audio signal power calculation means 36 is a means for calculating the magnitude of the audio signal that is relatively emphasized due to the noise being attenuated by the noise band selection attenuation means 18.

The S/N calculating means 38 is a means for inputting the audio signal calculated by the audio signal power calculating means 36 and the noise power calculated by the noise power calculating means 37, and calculating an S/N ratio. It is. Then, the band selection attenuation control means 17 inputs the S/N ratio calculated by the S/N ratio calculation means 38, and sends a control signal such that the S/N ratio becomes a predetermined target S/N value. is output to the audio band selection emphasis means 18.

In the embodiments described above, the means such as the voice band detection means and the voice band selection emphasis means can be realized in software using a computer, but it is also possible to use dedicated hardware circuits having the respective functions. is also possible.

As is clear from the explanation in the above embodiments, the audio signal processing device according to the present invention predicts the noise component and uses the cancellation coefficient when canceling the noise. Emphasizes the audio level of the band,
Alternatively, since the noise level in the noise band is attenuated, it is possible to obtain an audio signal with further suppressed noise.

[0081]

As explained above, according to the present invention, the band in which the voice is included is detected by the voice detection means among the band-divided frequency components of the voice signal including noise, and the band in which the voice signal is included is detected. By emphasizing the noise and providing a noise suppression system including this voice detection means, it is possible to improve the SNR of the voice signal containing noise.

It goes without saying that the speech recognition rate can be improved by installing the speech signal processing device of the present invention in front of the speech recognition device.

[0083] The present invention can also be applied to various communication systems to improve communication quality.

[Brief explanation of the drawing]

FIG. 1 A block diagram showing an embodiment of an audio signal processing device according to the present invention.

[Figure 2] Block diagram showing more specific details of the embodiment

[Figure 3] Block diagram showing another embodiment of the present invention

[
FIG. 4 Block diagram showing another embodiment of the present invention

[Figure 5
] Block diagram showing another embodiment of the present invention

[Figure 6]
Block diagram showing another embodiment of the present invention

[Fig. 7] Block diagram showing another embodiment of the present invention

[Fig. 8] A block diagram showing an embodiment of the audio signal processing device according to the present invention.

[Figure 9] Block diagram showing more specific details of the embodiment

FIG. 10: Block diagram showing another embodiment of the present invention

[Fig. 11] Block diagram showing another embodiment of the present invention

[
FIG. 12 Block diagram showing another embodiment of the present invention

[Fig. 13] Block diagram showing another embodiment of the present invention

[Figure 1
4] Block diagram showing another embodiment of the present invention

[Figure 15
] Block diagram showing another embodiment of the present invention

[Figure 16]
A block diagram showing an embodiment of an audio signal processing device according to the present invention

FIG. 17 A block diagram showing another embodiment of the audio signal processing device according to the present invention.

FIG. 18: Current status 1 of the audio signal processing device according to the present invention
Block diagram showing one embodiment

FIG. 19 is a block diagram showing another embodiment of the audio signal processing device according to the present invention.

FIG. 20 Graph for explaining cepstrum analysis in one embodiment of the present invention

[Figure 21] Graph for explaining the voice band and noise band in the present invention

[Figure 22] Graph for explaining the method of noise prediction in the present invention

[Figure 23] Graph for explaining the cancellation method in the present invention

[Figure 24] Graph for explaining the cancellation coefficient in the present invention

[Figure 25] Block diagram showing a conventional audio signal processing device

[Explanation of symbols]

11 Band division means 12 Audio band detection means 13 Band selection emphasis control means 14 Audio band selection emphasis means 15 Band synthesis means 16 Noise band calculation means 17 Band selection attenuation control means 18 Noise band selection attenuation means 19 Emphasis/attenuation means 21 Cepstral analysis means 22 Peak detection means 23 Audio band detection circuit 24 Formant analysis means 32 Voice discrimination circuit 33 Voice prediction means 34 Cancellation means 36 Audio signal power calculation means 37 Noise power calculation means 38 S/N calculation means 42 Pitch frequency detection means 43 Cancellation coefficient setting means

Claims (19)

[Claims] Claims: 1. A band dividing means for inputting a mixed signal containing noise and dividing the signal into bands; and an audio band for detecting an audio band portion of each band-divided signal for each channel by the band dividing means. a detection means and a voice band selection/emphasis means for relatively emphasizing the voice signal band with respect to the noise signal band for the band-divided noise-containing signal based on the voice band information detected by the voice band detection means; and a band synthesizing means for band-synthesizing the signal emphasized by the audio band selection emphasizing means. 2. A band dividing means for inputting a mixed signal containing noise and dividing the band into bands; and an audio band for detecting an audio band portion of each band-divided signal for each channel by the band dividing means. a detection means, a band selection emphasis control means for outputting a control signal for emphasizing a voice band based on the voice band information detected by the voice band detection means, and a signal mixed with noise band-divided by the band division means. is input, and in accordance with the control signal of the band selection emphasis control means, selects an audio band, and performs band synthesis on the audio band selection emphasis means that emphasizes only that band and the signal emphasized by the audio band selection emphasis means. An audio signal processing device comprising: band synthesis means. 3. The audio band detection means includes: cepstrum analysis means for inputting the band-divided signal and performing cepstrum analysis; peak detection means for detecting peaks based on the cepstrum analysis results; Claim 1, further comprising a voice band detection circuit that detects a voice band using the detected peak.
The audio signal processing device described above. 4. The audio band detection means includes: cepstrum analysis means for inputting the band-divided signal and performing cepstrum analysis; peak detection means for detecting a peak based on the cepstrum analysis results; a formant analysis means for performing formant analysis based on the
2. A voice band detection circuit according to claim 1, further comprising a voice band detection circuit that detects a voice band by using the formant information analyzed by the formant analysis means and the peak detected by the peak detection means. Audio signal processing device. 5. A band dividing means for inputting a mixed signal containing noise and dividing the band into bands; and an audio band for detecting an audio band portion of each band-divided signal for each channel by the band dividing means. A detection means, a noise band calculation means for calculating a noise band based on the voice band information detected by the voice band detection means, and a control signal for emphasizing the noise band calculated by the noise band calculation means. band selection attenuation control means; and noise band selection attenuation means which inputs the noise-mixed signal band-split by the band division means, selects a noise band according to a control signal of the band selection attenuation control means, and attenuates only that noise band. and band synthesis means for band-synthesizing the signal attenuated by the noise band selection attenuation means. 6. The audio band detection means includes: cepstrum analysis means for inputting the band-divided signal and performing cepstrum analysis; peak detection means for detecting a peak based on the cepstrum analysis result; a formant analysis means for performing formant analysis based on the
6. The voice band detection circuit according to claim 5, further comprising a voice band detection circuit that detects a voice band by using the formant information analyzed by the formant analysis means and the peak detected by the peak detection means. Audio signal processing device. 7. A band dividing means for inputting a mixed signal containing noise and dividing the band into bands; and an audio band for detecting an audio band portion of each band-divided signal for each channel by the band dividing means. a detection means; a band selection emphasis control means for outputting a control signal for emphasizing a voice band based on the voice band information detected by the voice band detection means; a noise band calculation means for calculating a noise band based on the noise band calculation means; a band selection attenuation control means for outputting a control signal for emphasizing the noise band calculated by the noise band calculation means; and a noise band divided by the band division means. Regarding the mixed signal, a voice band is selected and emphasized only in accordance with the control signal of the band selection emphasis control means, and a noise band is selected and only that band is attenuated in accordance with the control signal of the band selection attenuation control means. An audio signal processing device comprising: an emphasis/attenuation means; and a band synthesis means for performing band synthesis on a signal emphasized/attenuated by the emphasis/attenuation means. 8. A band dividing means for inputting a mixed signal containing noise and dividing the band into bands; a voice discriminating means for discriminating a voice part of the signal band-divided by the band dividing means; noise prediction means for predicting noise in the audio part using the audio part information obtained by the above-mentioned audio part; canceling means for subtracting the predicted noise value from the band-divided signal; audio band detection means for detecting an audio band portion of the signal;
voice band selection and emphasis means for relatively emphasizing the voice signal band with respect to the noise signal band for the signal whose noise has been canceled by the canceling means, based on the voice band information detected by the voice band detection means; An audio signal processing device comprising: band synthesis means for band-synthesizing the signal emphasized by the audio band selection emphasis means. 9. A band division means for inputting a mixed signal containing noise and dividing the band into bands; and a cepstrum analysis means for performing cepstral analysis on the signal of each channel divided into bands by the band division means. , a peak detection means for detecting a peak based on the cepstral analysis result, a speech discrimination circuit for discriminating a speech part using the peak detected by the peak detection means, and a speech discrimination circuit that uses the discriminated speech part information. noise prediction means for predicting the noise in the voice part; canceling means for subtracting the predicted noise value from the band-divided signal; a voice band detection circuit for detecting the voice band detection circuit; a band selection emphasis control means for outputting a control signal for emphasizing the voice band based on the voice band information detected by the voice band detection circuit; A voice band selection and emphasis means inputs a signal, selects a voice band and emphasizes only that voice band according to a control signal of the band selection emphasis control means, and performs band synthesis on the signal emphasized by the voice band selection and emphasis means. An audio signal processing device comprising a band synthesis means. 10. The speech recognition circuit further comprises formant analysis means for performing formant analysis on the cepstrum by the cepstrum analysis means, and the speech discrimination circuit also uses the formant analysis results to discriminate speech parts. The audio signal processing device according to claim 2. 11. Band dividing means for inputting a mixed signal containing noise and dividing the band into bands, and by the band dividing means,
A cepstrum analysis means that performs cepstrum analysis on each band-divided signal for each channel, a peak detection means that detects peaks based on the cepstrum analysis results, and a peak detected by the peak detection means. a speech discrimination circuit that discriminates the speech portion; a noise prediction means that predicts the noise of the speech portion using the discriminated speech portion information; A noise band is calculated based on a cancellation means for subtracting, a voice band detection circuit that detects a voice band using the peak detected by the peak detection means, and voice band information detected by the voice band detection circuit. a noise band calculating means for calculating the noise band; a band selection attenuation control means for outputting a control signal for attenuating the noise band calculated by the noise band calculating means; and a signal whose noise has been canceled by the canceling means. The noise band selection attenuation means selects a noise band and attenuates only that noise band in accordance with the control signal of the selection attenuation control means, and the band synthesis means performs band synthesis on the signals attenuated by the noise band selection attenuation means. An audio signal processing device characterized by: 12. The speech recognition circuit further comprises formant analysis means for performing formant analysis on the cepstrum by the cepstrum analysis means, and the speech discrimination circuit also uses the result of the formant analysis to discriminate speech parts. The audio signal processing device according to claim 4. 13. Band dividing means for inputting a mixed signal containing noise and dividing the band into bands, and by the band dividing means,
A cepstrum analysis means that performs cepstrum analysis on each band-divided signal for each channel, a peak detection means that detects peaks based on the cepstrum analysis results, and a peak detected by the peak detection means. a speech discrimination circuit that discriminates the speech portion; a noise prediction means that predicts the noise of the speech portion using the discriminated speech portion information; a voice band detection circuit that detects a voice band using the peak detected by the peak detection means; and a voice band detection circuit that emphasizes the voice band based on the voice band information detected by the voice band detection circuit. a band selection emphasis control means for outputting a control signal to control the noise band; a noise band calculation means for calculating a noise band based on the voice band information detected by the voice band detection circuit; and a noise band calculation means for calculating a noise band based on the voice band information detected by the voice band detection circuit; band selection attenuation control means for outputting a control signal for emphasizing a band; and for the signal whose noise has been canceled by the canceling means, selecting an audio band and emphasizing only it according to a control signal of the band selection emphasis control means; Further, according to the control signal of the band selection attenuation control means, a noise band is selected and only the noise band is attenuated. 1. An audio signal processing device comprising: a synthesizing means. 14. Band dividing means for inputting a mixed signal containing noise and dividing the band into bands, and by the band dividing means,
A cepstrum analysis means that performs cepstrum analysis on each band-divided signal for each channel, a peak detection means that detects peaks based on the cepstrum analysis results, and a peak detected by the peak detection means. a speech discrimination circuit that discriminates the speech portion; a noise prediction means that predicts the noise of the speech portion using the discriminated speech portion information; a voice band detection circuit that detects a voice band using the peak detected by the peak detection means; and a voice band detection circuit that emphasizes the voice band based on the voice band information detected by the voice band detection circuit. a band selection emphasis control means for outputting a control signal to make the noise, and a signal from which noise has been removed by the canceling means inputted, and according to the control signal of the band selection emphasis control means, selects an audio band and emphasizes only that audio band. It comprises an emphasizing means and a band synthesizing means for band-synthesizing the signal emphasized by the audio band selection emphasizing means, and further inputs the noise predicted by the noise predicting means and calculates the magnitude of the noise. a noise power calculation means for calculating the magnitude of the audio signal emphasized by the audio band selection and emphasis means; and an audio signal calculated by the audio signal power calculation means;
S/N ratio calculation means inputs the noise power calculated by the noise power calculation means and calculates the S/N ratio, and the band selection emphasis control means calculates the S/N ratio by the S/N ratio calculation means. Enter the S/N ratio that was
An audio signal processing device characterized by outputting a control signal such that the ratio becomes a predetermined target S/N value to the audio band selection emphasis means. 15. Band dividing means for inputting a mixed signal containing noise and dividing the band into bands, and by the band dividing means,
A cepstrum analysis means that performs cepstrum analysis on each band-divided signal for each channel, a peak detection means that detects peaks based on the cepstrum analysis results, and a peak detected by the peak detection means. a speech discrimination circuit that discriminates the speech portion; a noise prediction means that predicts the noise of the speech portion using the discriminated speech portion information; A noise band is calculated based on a cancellation means for subtracting, a voice band detection circuit that detects a voice band using the peak detected by the peak detection means, and voice band information detected by the voice band detection circuit. a noise band calculating means for calculating the noise band; a band selection attenuation control means for outputting a control signal for emphasizing the noise band calculated by the noise band calculating means; and a signal whose noise has been canceled by the canceling means. The noise band selection attenuation means selects a noise band and attenuates only that noise band in accordance with the control signal of the selection attenuation control means, and the band synthesis means performs band synthesis on the signals attenuated by the noise band selection attenuation means. Further, the noise power calculation means inputs the noise predicted by the noise prediction means and calculates the magnitude of the noise, and the noise band selection attenuation means calculates the magnitude of the relatively emphasized audio signal. an S/N ratio calculation means for inputting the audio signal calculated by the audio signal power calculation means and the noise power calculated by the noise power calculation means to calculate an S/N ratio; The band selection control means inputs the S/N ratio calculated by the S/N ratio calculation means, and generates a control signal such that the S/N ratio becomes a predetermined target S/N value. An audio signal processing device characterized in that the audio signal processing device outputs the signal to the noise band selection attenuation means. 16. Band division means for dividing a frequency band of an inputted audio signal containing noise; noise prediction means for receiving the band division output of the band division means and predicting its noise component; a pitch frequency detection means for detecting a pitch frequency of an input signal including a pitch frequency, a cancellation coefficient setting means for setting a cancellation coefficient according to a pitch frequency output of the pitch frequency detection means, a noise prediction output of the noise prediction means, and a noise prediction output of the noise prediction means; A canceling means that receives the band division output of the division means and the cancellation coefficient signal set by the cancellation coefficient setting means, and cancels a noise component from the band division output in consideration of the cancellation rate, and the pitch frequency detection means voice band detection means for detecting the voice band portion of the input signal using the pitch frequency detected by the voice band detection means; and band division selection emphasis control for outputting a control signal for emphasizing the voice band detected by the voice band detection means. a voice band selection and emphasis means for relatively emphasizing the voice signal band relative to the noise signal band for the signal whose noise has been canceled by the canceling means, according to a control signal of the band selection and emphasis control means; 1. An audio signal processing device comprising: band synthesizing means for band-synthesizing the signal emphasized by the audio band selection emphasizing means. 17. A band dividing means for dividing a frequency band of an input audio signal containing noise; a noise prediction means for receiving the band dividing output of the band dividing means and predicting its noise component; pitch frequency detection means for detecting the pitch frequency of the input signal including; cancellation coefficient setting means for setting a cancellation coefficient according to the pitch frequency output of the pitch frequency detection means; noise prediction output of the noise prediction means; and the band division. The band division output of the means and the cancellation coefficient signal set by the cancellation coefficient setting means are input, and from the band division output, canceling means cancels a noise component in consideration of the cancellation rate, and the pitch frequency detection means a voice band detection means for detecting a voice band portion of the input signal using the detected pitch frequency; and a noise band calculated based on voice band information detected by the voice band detection means. a noise band calculating means for calculating the noise band; a band selection attenuation control means for outputting a control signal for attenuating the noise band calculated by the noise band calculating means; and a signal whose noise has been canceled by the canceling means. The noise band selection attenuation means selects a noise band and attenuates only that noise band in accordance with the control signal of the selection attenuation control means, and the band synthesis means performs band synthesis on the signals attenuated by the noise band selection attenuation means. An audio signal processing device characterized by: 18. Band division means for dividing a frequency band of an input audio signal containing noise; noise prediction means for receiving the band division output of the band division means and predicting its noise component; a pitch frequency detection means for detecting a pitch frequency of an input signal including a pitch frequency, a cancellation coefficient setting means for setting a cancellation coefficient according to a pitch frequency output of the pitch frequency detection means, a noise prediction output of the noise prediction means, and a noise prediction output of the noise prediction means; A canceling means that receives the band division output of the division means and the cancellation coefficient signal set by the cancellation coefficient setting means, and cancels a noise component from the band division output in consideration of the cancellation rate, and the pitch frequency detection means voice band detection means for detecting the voice band portion of the input signal using the pitch frequency detected by the voice band detection means; and band selection emphasis control means for outputting a control signal for emphasizing the voice band detected by the voice band detection means. a voice band selection and emphasis means for relatively enhancing the voice signal band with respect to the noise signal band for the signal whose noise has been canceled by the cancellation means, according to a control signal of the band selection and emphasis control means; a band synthesizing means for band-synthesizing the signal emphasized by the band selection emphasizing means; further inputting noise predicted by the noise predicting means;
noise power calculation means for calculating the magnitude of noise; audio signal power calculation means for calculating the magnitude of the audio signal emphasized by the audio band selection and emphasis means; and audio calculated by the audio signal power calculation means. comprising an S/N ratio calculation means for inputting a signal and the noise power calculated by the noise power calculation means and calculating an S/N ratio,
The band selection emphasis control means inputs the S/N ratio calculated by the S/N ratio calculation means, and outputs a control signal such that the S/N ratio becomes a predetermined target S/N value. An audio signal processing device characterized by outputting to audio band selection emphasis means. 19. Band division means for dividing a frequency band of an input audio signal containing noise; noise prediction means for receiving the band division output of the band division means and predicting its noise component; a pitch frequency detection means for detecting a pitch frequency of an input signal including a pitch frequency, a cancellation coefficient setting means for setting a cancellation coefficient according to a pitch frequency output of the pitch frequency detection means, a noise prediction output of the noise prediction means, and a noise prediction output of the noise prediction means; A canceling means that receives the band division output of the division means and the cancellation coefficient signal set by the cancellation coefficient setting means, and cancels a noise component from the band division output in consideration of the cancellation rate, and the pitch frequency detection means audio band detection means for detecting an audio band portion of the input signal using the pitch frequency detected by;
Noise band calculation means for calculating a noise band based on the voice band information detected by the voice band detection means;
A band selection attenuation control means outputs a control signal for attenuating the noise band calculated by the noise band calculation means, and a signal whose noise has been canceled by the canceling means is input, and a control signal for the band selection attenuation control means is input. Accordingly, the noise band division selection attenuation means selects a noise band and attenuates only that noise band, and the band synthesis means performs band synthesis on the signal attenuated by the noise band selection attenuation means; noise power calculation means for inputting the noise predicted by the means and calculating the magnitude of the noise; and audio signal power calculation means for calculating the magnitude of the audio signal relatively emphasized by the noise band selection attenuation means. and S/N ratio calculation means for inputting the noise power calculated by the audio signal power calculation means and calculating the S/N ratio, and the band selection attenuation control means includes
Input the S/N ratio calculated by the N ratio calculation means,
An audio signal processing device characterized in that a control signal is output to the noise band selection attenuation means so that the S/N ratio becomes a predetermined target S/N value.