| 摘要 |
<P>PROBLEM TO BE SOLVED: To solve the problem in which an input speech can not be recognized correctly owing to circumferential noise, a noisy sound, etc., when inputted and speech input for control over controlled equipment is needed again. <P>SOLUTION: In registration mode, a control part 17 generates 1st data on movements of the lips of a user according to an image signal from a camera part 14, analyzes a 1st speech pattern according to a speech signal representing control contents inputted through a microphone 12, and stores them in a storage part 18 while relating them to data showing control contents inputted through an operation part 11. In control mode, the control part 17 recognizes data matching 2nd data on movements of the lips of the user based upon the image signal from the camera part 14 among 1st data in the storage part 18 in case of failure of speech recognition analysis for retrieving a 2nd speech pattern obtained by analyzing the speech inputted from the microphone 12 from 1st speech patterns in the storage part 18, thereby extracting related control contents. <P>COPYRIGHT: (C)2004,JPO&NCIPI |
