VOICE RECOGNITION DEVICE

PURPOSE:To improve the recognition rate of a continuous sound by extracting a syllable label which contains a phoneme label extracted from the syllable label of an extracted syllable recognition candidate as a syllable recognition result. CONSTITUTION:This device is provided with a 1st extracting me...

Full description

Saved in:
Bibliographic Details
Main Author FUJIMAKI SAKAE
Format Patent
LanguageEnglish
Published 01.05.1990
Subjects
ACOUSTICS
MUSICAL INSTRUMENTS
PHYSICS
SPEECH ANALYSIS OR SYNTHESIS
SPEECH OR AUDIO CODING OR DECODING
SPEECH OR VOICE PROCESSING
SPEECH RECOGNITION
Online AccessGet full text

Cover

More Information
Summary:PURPOSE:To improve the recognition rate of a continuous sound by extracting a syllable label which contains a phoneme label extracted from the syllable label of an extracted syllable recognition candidate as a syllable recognition result. CONSTITUTION:This device is provided with a 1st extracting means 3 which selects a syllable label of high similarity among syllable labels by referring a 1st standard feature parameter stored in a 1st storage means 100. A 2nd extracting means 400 selects the phoneme label of the highest similarity among phoneme labels by referring to a 2nd standard feature parameter stored in a 2nd storage means 200. Further, a 3rd extracting means 500 which extracts the syllable label containing the phoneme label extracted by the 2nd extracting means 400 among syllable recognition candidates extracted by the 1st extracting means 300 as a syllable recognition result. Consequently, even if misrecognition is caused, matching with the recognition result in phoneme units is performed, so the correct recognition candidate can be detected.
Bibliography:Application Number: JP19880269432