A method for Chinese syllables recognition based upon sub-syllable hidden Markov model

Author

Ji, Tianying ; Wang, Zuoying ; Lu, Dajin

Author_Institution

Dept. of Electron. Eng., Tsinghua Univ., Beijing, China

fYear

1994

fDate

13-16 Apr 1994

Firstpage

730

Abstract

All Chinese words are single syllables with C/V (Consonant/Vowel) structure. In this paper, a method for Chinese syllables recognition based upon the unique C/V structure of Chinese syllables is described. All the 1254 Chinese toned syllables (excluding some rare syllables) are decomposed into 155 sub-syllable units, 117 for consonants and 38 for vowels. In this way, the problem of performance degradation caused by insufficient training data is avoided, and the robustness of recognition is improved. Experimental results show that recognition rate is above 90% when applying this method to a real-time recognition system

Keywords

character sets; hidden Markov models; speech recognition; Chinese syllables recognition; consonant/vowel structure; real-time recognition system; recognition rate; robustness; sub-syllable hidden Markov model; sub-syllable units; toned syllables; Character recognition; Degradation; Hidden Markov models; Mice; Natural languages; Real time systems; Robustness; Speech recognition; Training data; Writing;

fLanguage

English

Publisher

ieee

Conference_Titel

Speech, Image Processing and Neural Networks, 1994. Proceedings, ISSIPNN '94., 1994 International Symposium on

Print_ISBN

0-7803-1865-X

Type

conf

DOI

10.1109/SIPNN.1994.344808

Filename

344808