Research on the non-uttering human interface between multiplelanguages using lips movements

我们发现，在没有可听话语的情况下，通过嘴唇运动识别单词的稳健性不仅存在于不同说话者的单词之间，而且存在于同一说话者在不同时间或不同日期的话语之间。此外，我们还展示了唇形运动对火车站名称识别的有效性。此外，无论母语是英语还是日语的人说话，都可以识别同一个英语单词。我们还开发了一个系统，该系统可以自动完成从发音时唇部动作的获取到所发出的单词的识别，并展示了在各种语言中非发音人机界面的可行性。

We found robustness of word recognition by lip movement without audible utterance, not only between the words of different speakers but also between the utterances of the same speaker at different times or on different days. Furthermore, we showed the effectiveness of lip movement for the recognition of the railway station names. Moreover, no matter whether a native English speaker or Japanese person spoke, recognition of the same English word was possible. We also developed a system which can beautomatically performed, from acquisition of lip movements when uttering the word to recognition of the uttered word, and showed the feasibility of the non-uttering human interface in various languages.