Title:
DEVICE AND METHOD FOR GATHERING SOUND
Document Type and Number:
Japanese Patent JP3537962
Kind Code:
B2
Abstract:
PROBLEM TO BE SOLVED: To extract all sounds from person's positions or only the sound from a specific person's position by determining a filter coefficient according to a person's position to be processed which is selected out of detected person's positions or a signal regarding the specific position and a sound source signal.
SOLUTION: A person's position detection part 2 detects the position of a person from an image inputted from an image input part 1. A noise suppression part 4 selects the person's position to be processed out of detected person's positions by a person's position selection part 4-1 and generates the learning signal of an adaptive filter according to the person's position selected by an environment adaptation part 4-2 to determine the coefficient of the adaptive filter. A noise removal part 4-3 filters an input sound from a sound input part 3 by using the determined filter coefficient to suppress its noise. Consequently, when plural persons speak at the same time, all the sounds of the persons can be extracted while having their background noise suppressed or the sound of one person can be extracted while the sounds of other persons are suppressed.
Inventors:
Nagata, Hitoshi
Application Number:
JP20621096A
Publication Date:
June 14, 2004
Filing Date:
August 05, 1996
Export Citation:
Assignee:
TOSHIBA CORP
International Classes:
H04R3/00; H03H17/00; H03H21/00; (IPC1-7): H04R3/00; H03H17/00; H03H21/00