Noise suppress processing apparatus and method

• US 6,339,758 B1
• Filed: 07/30/1999
• Issued: 01/15/2002
• Est. Priority Date: 07/31/1998
• Status: Expired due to Term

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Kanazawa, Hiroshi, Akamine, Masami
• Primary Examiner(s): Korzuch, William
• Assistant Examiner(s): Lerner, Martin
• Application Number:
  US09/363,843
• Time in Patent Office: 900 Days
• Field of Search: 704/226, 704/227, 704/228, 381/71.1, 381/71.11, 381/71.14, 381/94.1, 381/94.2, 381/94.3, 381/94.7, 381/94.9
• US Class Current: 704/226
• CPC Class Codes: G10L 2021/02166 Microphone arrays; Beamforming
• CPC Class Codes: G10L 21/02 Speech enhancement, e.g. noise reduction or echo cancellation reducing echo effects in line transmission systems H04B3/20; echo suppression in hands-free telephones H04M9/08

Abstract

A noise suppress processing apparatus has a speech input section for detecting speech uttered by the speaker at different positions, an analyzer section for obtaining frequency components in units of channels by frequency-analyzing speech signals in units of speech detecting positions, a first beam former processor section for obtaining target speech components by suppressing noise in the speaker direction by filtering the frequency components in units of channels using filter coefficients, which are calculated to decrease the sensitivity levels in directions other than a desired direction, a second beam former processor section for obtaining noise components by suppressing the speech of the speaker by filtering the frequency components for the plural channels obtained by the analyzer section to set low sensitivity levels in directions other than a desired direction, an estimating section for estimating the noise direction from the filter coefficients of the first beam former processor section, and estimating the target speech direction from filter coefficients of the second beam former processor section, and a correcting section for correcting a first input direction as the arrival direction of the target speech to be input in the first beam former processor section on the basis of the target speech direction estimated by the estimating section, and correcting a second input direction as the arrival direction of noise to be input in the second beam former processor section on the basis of the noise direction estimated by the estimating section.

Claims

1. A noise suppression apparatus for independently outputting speech frequency components and noise frequency components, comprising:

a speech input section which receives speech uttered by a speaker at different positions and generates speech signals corresponding to the different positions;

a frequency analyzer section which frequency-analyzes the speech signals in units of channels of the speech signals to output frequency components for a plurality of channels;

a first beam former processor section which suppresses arrival noise other than a target speech by adaptive filtering using the frequency components for the plurality of channels to output the target speech;

a second beam former processor section which suppresses the target speech by adaptive filtering using the frequency components for the plurality of channels to outputting noise;

a noise direction estimating section which estimates a noise direction from filter coefficients calculated by the first beam former processor section;

a target speech direction estimating section which estimates a target speech direction from filter coefficients calculated by said second beam former processor section;

a target speech direction correcting section which corrects a first input direction as an arrival direction of the target speech to be input in said first beam former processor section on the basis of the target speech direction estimated by said target speech direction estimating section; and

a noise direction correcting section which corrects a second input direction as an arrival direction of noise to be input in said second beam former processor section on the basis of the noise direction estimated by said noise direction estimating section.

2. An apparatus according to claim 1, further comprising a spectrum subtraction noise suppression section including a speech band power calculator section which divides the obtained speech frequency components in units of frequency bands and calculates speech power for each band, a noise band power calculator section which divides the obtained noise frequency components in units of frequency bands and calculates noise power for each band, and a spectrum subtractor section which suppresses background noise by weighting in units of frequency bands of speech signals on the basis of the speech and noise frequency band power values obtained by said speech and noise band power calculator sections.

3. An apparatus according to claim 1, further comprising a speech band power calculator section which divides the obtained speech frequency components in units of frequency bands and calculates speech power for each band;

a noise band power calculator section which divides the obtained noise frequency components in units of frequency bands and calculates noise power for each band;

an input band power calculator section which divides, in units of frequency bands, frequency components of input signals obtained by frequency-analyzing the input signals obtained from said speech input section and calculates input power for each band; and

a corrected spectrum subtractor section for suppressing background noise by weighting in units of frequency bands of speech signals on the basis of the input band power, speech band power, and noise band power.

4. An apparatus according to claim 1, wherein said frequency analyzer section converts the speech signal components for the plurality of channels in a time domain into signal components in a frequency domain by the fast Fourier transform, and outputs frequency spectrum data in units of channels.

5. An apparatus according to claim 1, wherein said target speech direction correcting section converts estimation amount information output from said target speech direction estimating section into angle information of a current target speech source direction, and outputs the angle information to said first beam former processor section.

6. An apparatus according to claim 1, wherein said noise direction correcting section converts estimation amount information output from said noise direction estimating section into angle information of a current target noise source direction, and outputs the angle information to said second beam former processor section.

7. An apparatus according to claim 1, wherein each of said first and second beam former processor sections comprises a phase shifter configured to set an input direction of the beam former processor section, and a beam former main section configured to suppress components from directions other than an arrival direction of signal components to be extracted.

8. An apparatus according to claim 1, wherein said speech input section has at least first and second microphones, which are placed at least two different positions, and output frequency components for at least two speech channels.

9. A noise suppression apparatus for independently outputting speech frequency components and noise frequency components, comprising:

a speech input section which receives speech uttered by a speaker at least at two different positions and generates speech signals corresponding to the speech receiving positions in units of channels;

a frequency analyzer section which frequency analyzes the speech signals and outputs frequency components for a plurality of channels;

a first beam former processor section which executes arrival noise suppression processing for suppressing speech components other than speech from a speaker direction to obtain a target speech component, the noise suppression processing being performed by adaptive filtering of the frequency components for the plurality of channels obtained by said frequency analyzer section, using filter coefficients which are calculated to decrease sensitivity levels in directions other than a desired direction;

a second beam former processor section which executes second speech suppression processing for suppressing the speech from the speaker direction to obtain a first noise component, the speech suppression processing being performed by adaptive filtering of the frequency components for the plurality of channels obtained by said frequency analyzer section, using filter coefficients which are calculated to decrease sensitivity levels in directions other than a desired direction;

a third beam former processor section which executes second speech suppression processing for suppressing the speech from the speaker direction to obtain a second noise component, the second speech suppression processing being performed by adaptive filtering of the frequency components for the plurality of channels obtained by said frequency analyzer section, using filter coefficients which are calculated to decrease sensitivity levels in directions other than a desired direction;

a noise direction estimating section which estimates a noise direction from the filter coefficients calculated by said first beam former processor section;

a first target speech direction estimating section which estimates a first target speech direction from the filter coefficients calculated by said second beam former processor section;

a second target speech direction estimating section which estimates a second target speech direction from the filter coefficients calculated by said third beam former processor section;

a first input direction correcting section which corrects a first input direction as an arrival direction of target speech to be input in said first beam former processor section on the basis of at least one of the first target speech direction estimated by said first target speech direction estimating section and the second target speech direction estimated by said second target speech direction estimating section;

a second input direction correcting section which, when the noise direction estimated by said noise direction estimating section falls with a predetermined first range, corrects a second input direction as an arrival direction of noise to be input in said second beam former processor section on the basis of the noise direction;

a third input direction correcting section which, when the noise direction estimated by said noise direction estimating section falls with a predetermined second range, corrects a second input direction as an arrival direction of noise to be input in said third beam former processor section on the basis of the noise direction; and

an effective noise determination section which determines one of the first and second output noise components as true noise output components on the basis of whether the noise direction estimated by said noise direction estimating section falls within the predetermined first or second ranges and outputs the determined output noise component, and at the same time, determines which estimation result of said first and second speech direction estimating sections is effective and outputs the determined speech direction estimation result to said first input direction correcting section.

10. An apparatus according to claim 9, further comprising a spectrum subtraction noise suppression section including a speech band power calculator section configured to divide the obtained speech frequency components in units of frequency bands and calculate speech power for each band, a noise band power calculator section configured to divides the obtained noise frequency components in units of frequency bands and calculates noise power for each band, and a spectrum subtractor section configured to suppress background noise by weighting in units of frequency bands of speech signals on the basis of the speech and noise frequency band power values obtained by said speech and noise band power calculator sections.

11. An apparatus according to claim 9, further comprising a speech band power calculator section configured to divide the obtained speech frequency components in units of frequency bands and calculate speech power for each band;

a noise band power calculator section configured to divide the obtained noise frequency components in units of frequency bands and calculate noise power for each band;

an input band power calculator section configured to divide, in units of frequency bands, frequency components of input signals obtained by frequency-analyzing the input signals obtained from said speech input section calculating input power for each band; and

a corrected spectrum subtractor section configured to suppress background noise by weighting in units of frequency bands of speech signals on the basis of the input band power, speech band power, and noise band power.

12. An apparatus according to claim 9, wherein said first input direction correcting section converts estimation amount information output from at least one of said first and second target speech direction estimating sections into angle information of a current target speech source direction, and outputs the angle information to said first beam former processor section.

13. An apparatus according to claim 9, wherein said second input direction correcting section converts estimation amount information output from said noise direction estimating section into angle information of a current target noise source direction, and outputs the angle information to said second beam former processor section.

14. An apparatus according to claim 9, wherein said third input direction correcting section converts estimation amount information output from said noise direction estimating section into angle information of a current target noise source direction, and outputs the angle information to said third beam former processor section.

15. A noise suppression method for independently outputting speech frequency components and noise frequency components, comprising the steps of:

receiving speech uttered by a speaker at different positions to obtain speech signals of different channels;

frequency-analyzing the speech signals in units of channels to obtain frequency spectrum components in units of channels;

suppressing arrival noise other than a target speech by adaptive filtering using the frequency spectrum components in units of channels obtained in the frequency analyzing step, to output the target speech;

suppressing the target speech by adaptive filtering using the frequency components in units of channels to obtain noise components;

estimating a noise direction from filter coefficients used in adaptive filtering and calculated in the step of suppressing arrival noise;

estimating a target speech direction from filter coefficients used in adaptive filtering and calculated in the step of suppressing the target speech;

correcting a first input direction as an arrival direction of the target speech to be input in the step of suppressing arrival noise on the basis of the target speech direction estimated in the step of estimating a target speech direction; and

correcting a second input direction as an arrival direction of noise to be input in the step of suppressing the target speech on the basis of the noise direction estimated by the step of estimating a noise direction.

16. A method according to claim 15, further comprising the steps of dividing the obtained speech frequency components in units of frequency bands, calculating speech power for each band, dividing the obtained noise frequency components in units of frequency bands, calculating noise power for each band, and suppressing background noise by weighting in units of frequency bands of speech signals on the basis of the speech and noise frequency band power values obtained in the speech and noise band power calculation steps.

17. A method according to claim 15, further comprising:

the steps of dividing the obtained speech frequency components in units of frequency bands, calculating speech power for each band, dividing the obtained noise frequency components in units of frequency bands, calculating noise power for each band, dividing frequency components of input signals obtained in the frequency analyzing step in units of frequency bands, calculating input power for each band, and suppressing background noise by weighting in units of frequency bands of speech signals on the basis of the input band power, speech band power, and noise band power.

18. A noise suppression method comprising the steps of:

receiving speech uttered by a speaker at different positions to obtain speech signals of different channels;

frequency-analyzing speech signals in units of channels to obtain frequency spectrum components in units of channels;

executing arrival noise suppression processing for suppressing speech components other than speech from a speaker direction to obtain target speech components, the arrival noise suppression processing being performed by adaptive filtering of the frequency spectrum components for the plurality of channels obtained in units of channels in the frequency analyzing step, using filter coefficients which are calculated to decrease sensitivity levels in directions other than a desired direction;

executing first speech suppression processing for suppressing the speech from the speaker direction to obtain first noise components, the first speech suppression processing being performed by adaptive filtering of the frequency components for the plurality of channels using the frequency components obtained in units of channels in the frequency analyzing step, using filter coefficients which are calculated to decrease sensitivity levels in directions other than a desired direction;

executing second speech suppression processing for suppressing the speech from the speaker direction to obtain first noise components, the second speech suppression processing being performed by adaptive filtering of the frequency spectrum components for the plurality of channels obtained in units of channels in the frequency analyzing step, using filter coefficients which are calculated to decrease sensitivity levels in directions other than a desired direction;

estimating a noise direction from the filter coefficients calculated in the step of suppressing arrival noise suppression processing;

estimating a first target speech direction from the filter coefficients calculated in the step of executing first speech suppression processing;

estimating a second target speech direction from the filter coefficients calculated in the step of executing second speech suppression processing;

correcting a first input direction as an arrival direction of target speech to be input in the step of executing arrival noise suppression processing on the basis of at least one of the first target speech direction and the second target speech direction;

correcting a second input direction as an arrival direction of noise to be input in the step of executing first suppression processing on the basis of the noise direction estimated in the noise direction estimating step, as needed, when the noise direction falls with a predetermined first range;

correcting a second input direction as an arrival direction of noise to be input in the step of executing second speech suppression processing on the basis of the noise direction, when the noise direction falls with a predetermined second range; and

determining one of the first and second output noise components as true noise output components on the basis of whether the noise direction estimated in the noise direction estimating step falls within the predetermined first or second ranges and outputting the determined output noise component, and at the same time, determining which estimation result in the first and second speech direction estimating steps is effective and outputting the determined speech direction estimation result as a speech direction estimation result to be used in the first input direction correcting step.

19. A method according to claim 18, further comprising the steps of dividing the obtained speech frequency components in units of frequency bands, calculating speech power for each band, dividing the obtained noise frequency components in units of frequency bands, calculating noise power for each band, and suppressing background noise by weighting in units of frequency bands of speech signals on the basis of the speech and noise frequency band power values obtained in the speech and noise band power calculation steps.

20. A method according to claim 18, further comprising the steps of by dividing the obtained speech frequency components in units of frequency bands, calculating speech power for each band, dividing the obtained noise frequency components in units of frequency bands calculating noise power for each band, dividing frequency components of input signals obtained in the frequency analyzing step in units of frequency bands calculating input power for each band, and the corrected spectrum subtraction step of suppressing background noise by weighting in units of frequency bands of speech signals on the basis of the input band power, speech band power, and noise band power.