| 摘要 |
<p>PURPOSE:To prevent a speaker from being erroneously recognized and shorten the response time by a method wherein the changing-over of category allotted to every user is performed together with the indication of speaker when key words are inputting in voice in an electronic range equipped with an voice recognition means. CONSTITUTION:A voice input part 18 amplifies the voice signal inputted from a microphone 8. A parameter comparison part 20 calculates a distance between characteristic parameter groups of various kinds of words registered in memories 21, 22 and 23, and the characteristic parameter of unknown voice inputted from a characteristic parameter extraction part 19 in order to output the result of recognition or a reject signal. A speaker indicating means 24 indicates the speaker in response to a speaker designation signal MA or MB. A category selecting means 25 selects only memories 21 and 22 when the MA signal is outputted, while selects memories 22 and 23 when the MB signal is outputted. Because the key word memory 22 is commonly used as just described above, the changing-over of speaker designation can be facilitated.</p> |
