| 摘要 |
PROBLEM TO BE SOLVED: To solve the problem of conventional reverberation removal of regarding reverberation attenuation characteristics at the end of utterance important with assumption of sensing termination of correct utterance despite the fact that utterance termination sensing operation is easily mistaken and the reverberation attenuation characteristics of a chamber, which should be constant in principal, are estimated differently by utterance due to absence of a model of the reverberation attenuation characteristics of the chamber to be used. SOLUTION: A reverberation removing device includes a weighting coefficient producing means of estimating a weighting coefficient for estimating a ratio that sound energy output from a sound source before an observation point occupies in the observed sound energy based on given reverberation time, a power spectrum memory means for storing the sound energy output before the observation point, and a reverberation component subtracting means for subtracting the energy output before the observation point, which is stored in the power spectrum memory means, from the observed sound energy using the weighting coefficient. COPYRIGHT: (C)2011,JPO&INPIT |
