| 摘要 |
PROBLEM TO BE SOLVED: To provide a target sound extracting apparatus capable of maintaining high noise removal performance, when a sound component corresponding to target sound is extracted from an acoustic signal obtained through a microphone by filtering out a noise component based on a two (2) input spectrum substraction processing, even if different noise arrives at the microphone from a plurality of directions or a position relation changes between a sound source of the target sound and the microphone. SOLUTION: The target sound extracting apparatus comprises a sound source position detection means 51 and 52 for detecting the position (distance) of a target sound source for each of a plurality of sub-microphones, a jamming signal formation portion 6 for performing weighting composite processing on a plurality of sub-acoustic signals Sg2<SB>1</SB>and Sg2<SB>2</SB>based on the detection result to form a jamming signal Sgn, and a target sound extraction portion 7 for performing a two (2) input spectrum substraction processing to extract an acoustic signal corresponding to target sound and to output the extracted signal, using subtraction processing of a time series feature vector in each of the main acoustic signal Sg1 and the jamming signal Sgn. COPYRIGHT: (C)2009,JPO&INPIT
|
