抄録
Environmental noise is an inevitable obstacle for any speech-signal proccesing-based systems. Although bone-conducted microphone, which is robust for external noise, seems “a panacea for this syndrome”, quality of bone-conducted speech is deteriorated due to suppression of its high frequency component. Therefore, to comfort bone-conducted speech communication, signal proccesing to ease the deteration is indispensable. This paper proposed a method to improve quality of bone-conducted speech by restoring its high frequency component. The authors compared spectrum envelop of bone-conducted speech with that of air-conducted speech with cepstrum analysis, and revealed that replacement of spectrum envelop of bone-conducted speech with that of air-conducted speech can compensate the high frequency component of bone-conducted speech. Consequently, to apply spectrum envelop replacement for each phonemes appropriately, the authors designed a system using codebook-mapping which corresponds to given bone-conducted speech by each phonemes. Prototype of proposed speech enhancer was evaluated with cepstrum distance and hearing test. The evaluation ensured the prototype is enable to enhance bone-conducted speech especially to voiced speech. The proposed method may contributes to realize applications using speech signal under various environments.
