| 摘要 |
PURPOSE: To inform a speaker of misdetection and enable reregistration by indicating which voice section the misdetection is caused in when the difference in length between voice sections detected by two feature quantity extraction part is larger than a predetermined threshold value. CONSTITUTION: A voice for standard pattern generation when inputted to a microphone 11 is converted into an electric signal, which is inputted to the feature quantity extraction parts 12 and 13 respectively. Extracted feature quantities while inputted to standard pattern storage parts 14 and 15 are inputted to a voice section decision part 16. The voice section decision part 16 compares the lengths of the voice sections extracted by the feature quantity extraction parts 12 and 13 with each other and a display part 17 informs the speaker that there is an error in the detection of the speech section when the voice sections have a difference larger than the predetermined threshold value. Consequently, if the error in the voice section detection is caused at the time of standard pattern registration, the speaker is informed of the error and the reregistration is enabled. |
