voice recognition device

Abstract

Abstract PURPOSE:To improve a voice recognition rate and speed by recognizing the information obtained via a voice process and the correlation between the generated sounds and the lip movements with use of an image processing method. CONSTITUTION:Only the lip part is separated/extracted out of an image fetched from a camera via a lip image separating action 5, and the lip form change is analyzed by an image processing part 6. Then a lip form data train obtained by retrieving the contents of a data base 4 containing the information on the lip form produced by a wire frame process as the data is sent to a recognition process part 8. The part 8 retrieves the contents of a data base 10 containing the correlation data between the generated sounds and the lip movements as well as the contents of a voice data base. Then the part 8 recognizes the phonemes as well as the symbolization data on the voice features obtained via a feature extracting part 7. Thus, the voice recognition rate and speed can be improved.

Inventors

• 👤 Shunji Handa
• 👤 伴田　俊二

Bibliographic Details