Japanese voice recognition equipment

Abstract

Abstract PURPOSE: To efficiently output paragraph candidates by outputting the homonyum with a top priority having the same patterns with the accent patterns of inputted voices while selecting paragraph candidates by combining the syllables obtained by recognizing the inputted Japanese speech in a syllable unit. CONSTITUTION: Voices inputted form a microphone are inputted to an A/D converter 2 through an amplifier 1. Feature values for every frame are extracted from the converted digital signals by a speech feature extracting section 3. Then, the matching between the segmented syllables and the standard speech pattern groups registered in a standard pattern memory 51 is conducted by a syllable pattern matching section 5 which is a syllable discrimination section and stored in a syllable lattice buffer 8. Then, a syllable fundamental frequency is calculated from the fundamental frequency sequence for every syllable segment frame extracted by a fundamental frequency (pitch) extraction 31 in an extraction section 3 and transmitted to a paragraph accent pattern buffer 9. Then, a collating is conducted by a paragraph dictionary 10 and an accent pattern dictionary 11 and if matching is obtained, a desired paragraph candidate is stored in a buffer 13.

Inventors

• 👤 Fumio Togawa
• 👤 Toru Ueda

Bibliographic Details