8th European Conference on Speech Communication and Technology

Geneva, Switzerland
September 1-4, 2003


Speech Recognition Using EMG; Mime Speech Recognition

Hiroyuki Manabe, Akira Hiraiwa, Toshiaki Sugimura

NTT DoCoMo Inc., Japan

The cellular phone offers significant benefits but causes several social problems. One such problem is phone use in places where people should not speak, such as trains and libraries. A communication style that would not require voiced speech has the potential to solve this problem. Speech recognition based on electromyography (EMG), which we call "Mime Speech Recognition" is proposed. It not only eases communication in socially sensitive environments, but also improves speech recognition accuracy in noisy environments. In this paper, we report that EMG yields stable and accurate recognition of 5 Japanese vowels uttered statically without generating voice. Moreover, the ability of EMG to handle consonants is described, and the feasibility of basing comprehensive speech recognition systems on EMG is shown.

Full Paper

Bibliographic reference.  Manabe, Hiroyuki / Hiraiwa, Akira / Sugimura, Toshiaki (2003): "Speech recognition using EMG; mime speech recognition", In EUROSPEECH-2003, 2717-2720.