MEMS Technologies for Ubiquitous Computing World: Study on Voice Recognition Utilizing Bone-Conducted Voice
MEMS Technologies for Ubiquitous Computing World: Study on Voice Recognition Utilizing Bone-Conducted Voice
复制标题
面向普适计算世界的 MEMS 技术：利用骨传导语音进行语音识别的研究
DOI：
10.1541/ieejsmas.124.272
复制
发表时间：
2004
影响因子：
--
通讯作者：
S. Nakagawa
中科院分区：
文献类型：
--
作者：
Yoshiyasu Yamada;Yoshimasa Hijikata;N. Kawahara;Y. Fujisaka;S. Nakagawa
利用骨导语音，这是非常强大的噪声环境下，语音识别进行了研究。利用LPC倒谱距离（LCD）和语音传输指数（STI）评价了在人头部多个测量点记录的骨导语音的语音识别和语音通信的适用性。实验结果表明，前额和面颊是骨导语音较好的采集点。为了提高骨导语音的质量，研制了一种具有倾斜频率响应的新型骨导语音传感器。使用该传感器，获得的骨导语音具有与气导语音相似的频率响应。提出了一种提高骨导语音识别率的谱减法。
Voice recognition utilizing bone-conducted voice, which is very robust under noisy environment, has been investigated. The suitability for voice recognition and voice communication of bone-conducted voice recorded at several measurement points on human head was evaluated using LPC cepstrum distance (LCD) and Speech Transmission Index (STI). Experimental results show that forehead and cheek are good as the acquisition points of bone-conducted voice. In order to improve the quality of bone-conducted voice, a novel sensor, which has an inclined frequency response, was fabricated. Using the sensor, acquired bone-conducted voice has a similar frequency response as that of air-conducted voice. Spectral Subtraction is suggested as a method to improve the voice recognition ratio of bone-conducted voice.