| 摘要 |
<P>PROBLEM TO BE SOLVED: To perform a word-spotting lip reading for recognizing keywords from a text targetting an utterance scene of the text which has not been targetted in the lip reading technology until now, by applying the spotting recognition technology not to speech information but to image information. <P>SOLUTION: Image pickup means 1 acquires a face image including at least a lip region. Region extraction means 2 extracts the lip region from the acquired face image. Feature quantity measuring means 3 measures a shape feature quantity from the extracted lip region. A measured feature quantity of a keyword utterance scene is registered in a keyword DB 4. Determination means 5 compares the measured feature quantity with the registered feature quantity of the keyword DB and determines the utterance content of the lips. Result display means 6 is a display device displaying it to a user. <P>COPYRIGHT: (C)2012,JPO&INPIT |
