| 摘要 |
PROBLEM TO BE SOLVED: To provide an acoustic coupling amount calculation technique capable of highly accurately deducing the acoustic coupling amount, regardless of the speech state and speedily following variation of an echo path, even when there is variation. SOLUTION: A reproduced signal x(k) is converted into a reproduced signal spectrum X<SB>L,ω</SB>for each frame (61), and a sound collection signal y(k) is converted into a sound collection signal spectrum Y<SB>L,ω</SB>for each frame (62). Next, regarding the frequency axis direction and the time axis direction, a cross spectrum expected value E<SB>t, f</SB>[X<SP>*</SP><SB>L,ω</SB>×Y<SB>L,ω</SB>] is determined by Equation (7) (by section 111). Furthermore, regarding the frequency axis direction, a power spectrum expected value E<SB>t, f</SB>[¾X<SB>L,ω</SB>¾<SP>2</SP>] is determined by a specific expression (by section 121). The estimated value ¾H<SB>L,ω</SB>¾<SP>2</SP>of the acoustic coupling amount is then obtained by a specific expression (by section 131). COPYRIGHT: (C)2008,JPO&INPIT
|
