摘要 |
PROBLEM TO BE SOLVED: To provide a word recognition device capable of recognizing a word by use of an image recognition technique without depending on a voice itself in order to precisely recognize the word even in a noisy environment. SOLUTION: When the lips of a speaker are image processed for recognition, images including the lips are taken every unit time, the difference between the images by the lapse of time is calculated, whereby the sounding state of one sound is detected, the motion of the lips is recognized as a character string composed of vowel sounds and "N" , and this character string is compared with a word dictionary storing a plurality of word data to be recognized as character strings of vowel sounds and "N" by matching, whereby the word is recognized. COPYRIGHT: (C)2005,JPO&NCIPI
|