| 摘要 |
PROBLEM TO BE SOLVED: To provide a speech recognition device, a speech recognition method and a robot device, by which natural and smooth interaction is attained. SOLUTION: Based on a speech feature amount of a speaker calculated based on a speech signal which is input to a speech input section, a response timing for outputting response sound from a loud speaker during interaction with the speaker, is assumed. When an assumption result is obtained that there is the response timing for outputting the response sound from the loud speaker during interaction with the speaker, it is determined whether or not, the response sound is output based on power just before the response timing. Thereby, the timing when the interaction partner may respond is assumed as the response timing, by assuming the response timing based on the speech feature amount which is a guideline for the interaction partner to respond. Moreover, by two step determination, even if there is incorrectly detected response timing at first time, output of unnatural response sound which is incorrectly assumed is avoided by discarding the response timing, and natural and smooth interaction is attained. COPYRIGHT: (C)2009,JPO&INPIT
|
