Method for localizing sources of signals in reverberant environments using sparse optimization

Patent metadata

Field	Value
Publication number	US-9251436-B2
Application number	US-201313776850-A
Country	US
Kind code	B2
Filing date	Feb 26, 2013
Priority date	Feb 26, 2013
Publication date	Feb 2, 2016
Grant date	Feb 2, 2016

Abstract

Source signals emitted in a reverberant environment from different locations are processed by first receiving input signals corresponding to the source signals by a set of sensors. Then, a sparsity-based support estimation is applied to the input signals according to a reverberation model to produce estimates of the source signals and locations of a set of sources emitting the source signals.

First claim

We claim:

1. A method for processing source signals, comprising the steps of:
emitted, by a set of sources at different locations in a reverberant environment, the source signals, wherein the source signals are acoustic signals;
acquiring, using a set of sensors, input signals corresponding to the source signals, wherein the sensors are microphones and the input signal at each frequency is modeled as a linear combination of all the source signals at that particular frequency, and wherein coefficients in the linear combination correspond to a frequency response of the environment from each location at that frequency; and
applying a sparsity-based support estimation to the input signals according to a reverberation model to produce estimates of the source signals and the different locations of the set of sources, wherein the applying is performed in a processor.

2. The method of claim 1 , wherein the input signals are in a form of discrete Fourier transforms.

3. The method of claim 1 , wherein a number of the sources is different than a number of the sensors.

4. The method of claim 1 , further comprising:
discretizing the environment into a grid of locations, wherein each locations is a potential source emitting one of the source signals.

5. The method of claim 4 , wherein a number of the sources is less than a number of the locations in the grid.

6. The method of claim 4 , further comprising:
enforcing a model-based truncation to enforce constraints on distances between the locations.

7. The method of claim 1 , wherein the input signals are convolutions of the source signals with impulse responses of the environment.

8. The method of claim 1 , wherein a number of the sources is unknown.

9. The method of claim 1 , wherein a sparsity pattern for all of the frequencies is identical.

10. The method of claim 1 , wherein the applying uses a total energy of the input signal at each location for all the frequencies.

11. The method of claim 1 , wherein the estimation is a joint-sparsity Compressive Sampling Matching Pursuit (CoSaMP) procedure.

12. The method of claim 1 , wherein the estimation is a joint-sparsity greedy procedure.

13. The method of claim 1 , wherein the estimation is a joint-sparsity convex procedure.

14. The method of claim 1 , further comprising:
determining a gradient at each frequency according to linear systems derived from frequency transforms of the impulse responses.

15. The method of claim 14 , wherein the applying comprises the steps of:
determining a proxy using the gradients;
identifying a support candidate using the proxy;
inverting the linear systems over the support candidate at each frequency;
determining a final support using an output of the inverting;
truncating and updating the output of the inverting; and
iterating until convergence.

16. The method of claim 14 wherein the linear systems are normalized.

Assignees

• Mitsubishi Electric Res Lab

Inventors

• Boufounos Petros T
• Le Roux Jonathan
• Kang Kang
• Hershey John R

Classifications

• G01S3/802 Primary

  Systems for determining direction or deviation from predetermined direction · CPC title

• G06F18/21345

  enforcing sparsity or involving a domain transformation · CPC title

• G06F18/2136

  based on sparsity criteria, e.g. with an overcomplete basis · CPC title

• G06K9/6244 Primary

  Physics · mapped topic

• G06K9/6249

  Physics · mapped topic