摘要 |
PURPOSE:To increase a processing speed by collating in detail an expected word to be recognised, after being selected, with an unknown input voice. CONSTITUTION:Maximum-point/minimum-point parameter extraction circuit part 6 extracts as point feature quantities feature quantities when the feature quantity of an input voice signal is as its maximum or minimum. The information on the point feature quantity supplied from circuit part 6 is sent to collation-expectant selection part 10, where it is collated with the reference point feature quantity registered previously so as to select a recognition-expectant word, and the unknown input voice is collated in detail with the determined recognition-expectant word. Consequently, a processing speed can be increased. |