Apparatus and method for tracking locations of plurality of sound sources

Patent metadata

Field	Value
Publication number	US-9264806-B2
Application number	US-201213665143-A
Country	US
Kind code	B2
Filing date	Oct 31, 2012
Priority date	Nov 1, 2011
Publication date	Feb 16, 2016
Grant date	Feb 16, 2016

Abstract

Disclosed are an apparatus and a method for tracking locations of a plurality of sound sources. According to the apparatus and the method, a task for searching sound source candidates is repeated at respective predetermined frames of microphone signals to collect sound source candidates, and only the collected sound source candidates are verified through beamforming, thereby more rapidly and accurately tracking the plurality of sound sources in spite of using a small number of microphones.

First claim

What is claimed is:

1. An apparatus for tracking locations of a plurality of sound sources, the apparatus comprising:
a microphone array comprising a plurality of linearly disposed microphones; and
a sound source candidate extractor, comprising a processor, to extract sound source candidates at respective predetermined frames from microphone signals received from the microphone array; and
a sound source candidate verifier to perform beamforminq on the sound source candidates extracted by the sound source candidate extractor, to select the plurality of sound source candidates having a predetermined value or higher of signal intensity from the sound source candidates obtained as a result of the beamforminq and to predict locations of actual sound sources based on the selected sound source candidates,
wherein the sound source candidate extractor comprises:
a sound source feature extractor to extract voice features required for tracking locations of sound sources at respective frames from microphone signals received from the microphone array; and
a sound source candidate group extractor to extract the sound source candidates, based on the sound source features extracted by the sound source feature extractor, and to extract a plurality of sound source candidate groups, each including sound source candidates having the same sound source direction, from the extracted sound source candidates.

2. The apparatus according to claim 1 , wherein each predetermined frame has a data volume of the microphone signal of 256, 512 or 1024 bits.

3. The apparatus according to claim 1 , wherein the sound source candidate extractor transforms the microphone signals through windowing and a fast fourier transform (FFT), extracts voice features via a predetermined algorithm, and extracts the plurality of sound source candidates based on the extracted voice features, wherein sound source candidates in frames having sound source features are assigned a predetermined sound source candidate value other than zero, while sound source candidates in frames having no sound source features are assigned a sound source candidate value of zero, and only the sound source candidates having the sound source candidate value are extracted as candidates at each frame.

4. A method for predicting locations of a plurality of sound sources, the method comprising:
receiving microphone signals from a microphone array comprising a plurality of linearly disposed microphones;
extracting sound source candidates at respective predetermined frames of the received microphone signals;
beamforming the extracted sound source candidates;
selecting sound source candidates having a predetermined value or higher of signal intensity using results of the beamforming; and
predicting locations of actual sound sources based on the selected sound source candidates,
wherein the extracting of sound source candidates comprises:
extracting sound source features at respective predetermined frames of the received microphone signals;
extracting sound source candidates based on the extracted sound source features; and
extracting a plurality of sound source candidate groups, each including sound source candidates having the same sound source direction, from the respective extracted sound source candidates.

5. The method according to claim 4 , wherein, during extraction of sound source candidates, each predetermined frame has a data volume of the microphone signal of 256, 512 or 1024 bits.

6. The method according to claim 4 , wherein the extracting sound source features at respective predetermined frames of the received microphone signals further comprises:
transforming the microphone signals through windowing and a fast fourier transform (FFT);
extracting voice features via a predetermined algorithm; and
extracting sound source candidates based on the extracted voice features, wherein sound source candidates in frames having sound source features are assigned a predetermined sound source candidate value other than zero, while sound source candidates in frames having no sound source features are assigned a sound source candidate value of zero, and only sound source candidates having the sound source candidate value are extracted as candidates at each frame.

7. The method according to claim 4 , wherein in the selecting of the sound source candidates further comprises:
selecting sound source candidates exceeding a predetermined signal intensity from the verified sound source candidates obtained as a result of the beamforming; and
predicting locations of actual sound sources based on the selected sound source candidates.

8. At least one non-transitory medium comprising computer readable code to control at least one processor to implement the method of claim 4 .

Assignees

• Samsung Electronics Co Ltd

Inventors

• Hyun Kyung Hak
• Kim Jin Sung

Classifications

• H04R3/005 Primary

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

• H04R2430/23

  Direction finding using a sum-delay beam-former · CPC title

• H04R2201/403

  Linear arrays of transducers · CPC title

• G01S3/8083

  determining direction of source · CPC title

• G01S3/80 Primary

  using ultrasonic, sonic or infrasonic waves · CPC title