A Design of Headset-Microphone-Array for Noisy Speech Capture and Recognition
Various noises and reverberations are in the depository and the factory, because of operating various annoying machines. The workers in such environments communicated and recognized after capturing high-quality target speech with closed microphones such as headset-microphones. The directional microphone was usually employed as conventional headset-microphones. It can form higher sensitivity in front of microphone and lower sensitivity at the back of one. Therefore, it can capture target speech with high-quality by steering higher sensitivity to human mouth and lower sensitivity to the other. However, it is difficult to sufficiently suppress the various noises and reverberations with conventional headset-microphones based on directional microphone in the depository and the factory environments. The workers thus can not smoothly communicate each other. To overcome this problem, we newly designed a headset-microphone system based on microphone array (we named “Headset-Microphone-Array), and we try to acquire the target speech with high-quality in various noises and reverberations environments such as the depository and the factory. As a result of evaluation experiment, we confirmed that speech recognition performance was improved 13% in directional-noisy environment, 17% in diffused-noisy environment and totally 15% compared with conventional headset-microphone array (SENHEISER HMD-25).
Naoya KAWABATA Takanobu NISHIURA Yuki DENDA Masaomi IIDA
Graduate School of Science and Engineering,Ritsumeikan University,Japan Research & Development Division,Murata Machinery,Ltd.,Japan
国际会议
The 10th Western Pacific Acoustics Conference(第十届西太平洋声学会议WESPAC X)
北京
英文
1-5
2009-09-21（万方平台首次上网日期，不代表论文的发表时间）