Extraction of reverberant sound using microphone arrays

Patent metadata

Field	Value
Publication number	US-2016293179-A1
Application number	US-201615178530-A
Country	US
Kind code	A1
Filing date	Jun 9, 2016
Priority date	Dec 11, 2013
Publication date	Oct 6, 2016
Grant date	—

Abstract

A method includes estimating a spatial coherence between a first diffuse sound portion in a first microphone signal and a second diffuse sound portion in a second microphone signal. The first microphone signal is captured by a first microphone and the second microphone signal is captured by a second microphone which is spaced apart from the first microphone in a known manner. The method further includes defining a linear constraint for filter coefficients of a diffuse sound filter, the linear constraint being based on the spatial coherence. The method also includes calculating at least one of signal statistics and noise statistics over the first microphone signal and the second microphone signal. The method also includes determining the filter coefficients of the diffuse sound filter by solving an optimization problem concerning at least one of the signal statistics and noise statistics while considering the linear constraint for the filter coefficients.

First claim

1 . A method comprising:
defining a linear constraint for filter coefficients of a diffuse sound filter, the linear constraint being based on a spatial coherence between a first diffuse sound portion in a first microphone signal and a second diffuse sound portion in a second microphone signal, the first microphone signal being captured by a first microphone and the second microphone signal being captured by a second microphone spaced apart from the first microphone in a known manner; calculating at least one of a direction of arrival of at least one direct sound, signal statistics over the first and second microphone signals, and noise statistics over the first and second microphone signals; and determining the filter coefficients of the diffuse sound filter by solving an optimization problem concerning at least one of the direction of arrival of the at least one direct sound, the signal statistics, and the noise statistics while considering the linear constraint for the filter coefficients.

2 . The method according to claim 1 , further comprising
providing the spatial coherence on the basis of a relative transfer function or a correlation of the diffuse sound between the first microphone and the second microphone.

3 . The method according to claim 1 ,
wherein the spatial coherence is based on a prior measurement of a relative transfer function or a correlation of the diffuse sound for a given environment during time periods in which no direct sound is present in the environment.

4 . The method according to claim 1 ,
wherein the spatial coherence is based on a theoretical relation for the diffuse sound, wherein a corresponding assumed diffuse sound field has assumed theoretical properties regarding a correlation of the diffuse sound between the first microphone and the second microphone.

5 . The method according to claim 1 , wherein the optimization problem is expressed by

w
m



(

k
,
n

)

=

argmin
w





J


(
w
)

subject to the linear constraint
w H b m ( k,n )=1,

wherein
w(k,n) is a vector of the filter coefficients of the diffuse sound filter;
w m (k,n) is a solution of the optimization problem based on an evaluation of a microphone signal at the m-th microphone;
J(w) is a cost function;
b m (k,n) is a vector of estimated spatial coherences, wherein the m′-th element of the vector is an estimated spatial coherence of the diffuse sound between the m-th microphone and the m′-th microphone;
k is a frequency domain index; and
n is a time domain index.

6 . The method according to claim 5 , wherein the cost function J(w) is based on one of noise statistics, a noise power spectral density (PSD) matrix, signal statistics, or a microphone power spectral density (PSD) matrix.

7 . The method according to claim 1 , further comprising
estimating at least one of a direction of arrival of at least one direct sound or a relative transfer function of the at least one direct sound between the first microphone and the second microphone; computing at least one direct sound constraint using the direction of arrival or the relative transfer function of the at least one direct sound, wherein the at least one direct sound constraint results in a suppression of the at least one direct sound.

8 . The method according to claim 1 , wherein a solution to the optimization problem is

w
m



(

k
,
n

)

=

φ
d



(

k
,
n

)

β
+
1



Φ
x

-
1



(

k
,
n

)



b
m

with
β=α(φ d b m H Φ x −1 b m )

wherein
w m (k,n) is a solution of the optimization problem based on an evaluation of a microphone signal at the m-th microphone;
b m (k,n) is a vector of estimated spatial coherences, wherein the m′-th element of the vector is an estimated spatial coherence of the diffuse sound between the m-th microphone and the m′-th microphone;
αΣ[0,1] is a user-defined control parameter by which the diffuse sound filter can be scaled between a minimum mean square error spatial filter and a filter that minimizes the output power while satisfying the diffuse sound constraint;
φ d is a diffuse sound power; and
Φ x is a power spectrum matrix of the microphone signals.

9 . The method according to claim 8 , further comprising estimating the diffuse sound power φ d on the basis of an auxiliary diffuse sound filter.

10 . The method according to claim 9 , wherein estimating the diffuse sound power φ d is performed based on

φ
^

d



(

k
,
n

)

=

w
1
H



(

k
,
n

)



Assignees

• Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E V

Inventors

• Thiergart Oliver
• Habets Emanuel

Classifications

• G10L21/0264 Primary

  characterised by the type of parameter measurement, e.g. correlation techniques, zero crossing techniques or predictive techniques · CPC title

• H04R3/005 Primary

  for combining the signals of two or more microphones (specially adapted for hearing aids H04R25/407) · CPC title

• G10L2021/02082

  the noise being echo, reverberation of the speech · CPC title

• G10L21/02

  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) · CPC title

• G10L21/0208

  Noise filtering · CPC title