| 摘要 |
PROBLEM TO BE SOLVED: To provide a sound source separating method and a system therefor that can precisely separate a target sound from a plurality of noises. SOLUTION: When the target sound S<SB>C</SB>and noises S<SB>R</SB>and S<SB>L</SB>are separated, a sound insulator 20 is arranged at a border position between a 1st space and a 2nd space sectioned by a plane containing the arrival direction of the target sound S<SB>C</SB>and 1st and 2nd sound receiving devices 30 and 40 provided on the 1st- and 2nd-space sides of the sound insulator have directivity in four directions and receive sounds; after frequencies are analyzed, a noise S<SB>R</SB>of the 1st-space side and other sounds (S<SB>C</SB>, S<SB>L</SB>) are separated and, a noise S<SB>L</SB>of the 2nd-space side and other sounds (S<SB>C</SB>, S<SB>R</SB>) are separated through 1st layer processing (for example, SAFIA) and further the target sound S<SB>C</SB>is separated through 2nd layer processing (for example, SS). Further, spectrum integration processing is preferably performed as 3rd layer processing. COPYRIGHT: (C)2005,JPO&NCIPI |
