Identification of Utterance Content Using Lip Movement Features

Automatic minute preparation systems are used to improve the efficiencies of meetings and operations, and they use various voice recognition. However, preparation for usage environment should be performed based on the number of conference participants. In addition, speech recognition accuracy decreases because of the duplication of environmental sounds other than speech. However, lip movements have characteristics that are unique to the uttered contents and can be acquired even in noisy environments. In this study, a lip movement feature for estimating utterances was analyzed, with the aim of improving the speech recognition accuracy of an automatic minute creation system. Time-series changes in the lip height and width and luminance values of the mouth area associated with speech were used for vowel identification. The vowel identification experiment indicated that the lip height and luminance value of the mouth area are useful for identification of the vowels "u" and "i," respectively.