A fully parallel Mandarin speech recognition system with very large vocabulary and almost unlimited texts

Author

Lee, Lin-shan ; Tseng, Chiu-Yu ; Lin, Yueh Hong ; Lee, Yumin ; Tu, S.L. ; Gu, H.Y. ; Liu, F.H. ; Chang, C.H. ; Hsieh, S.H. ; Chen, C.H. ; Huang, K.R.

Author_Institution

Dept. of Comput. Sci. & Inf. Eng., Nat. Taiwan Univ., Taiwan

fYear

1991

fDate

11-14 Jun 1991

Firstpage

578

Abstract

The authors describe a fully parallel real-time Mandarin dictation machine which recognizes Mandarin speech with almost unlimited texts and a very large vocabulary for the input of Chinese characters to computers. Isolated syllables including the tones are first recognized using specially trained hidden Markov models with special feature parameters, and the exact characters are then identified from the syllables using a Markov Chinese language model, because every syllable can represent many different homonym characters. The real-time implementation is in Occam language on a transputer system with 10 T800 processors operating in parallel. The overall correction rate for the final output characters is about 80%

Keywords

computational linguistics; natural languages; neural nets; speech recognition; transputers; Mandarin dictation machine; Mandarin speech recognition system; Markov Chinese language model; Occam language; T800 processors; computer input; correction rate; exact characters; feature parameters; final output characters; homonym characters; input of Chinese characters; isolated syllables; large vocabulary; parallel speech recognition system; real-time implementation; tones; trained hidden Markov models; transputer system; unlimited texts; Character recognition; Computer science; Educational institutions; Hidden Markov models; History; Natural languages; Real time systems; Speech recognition; Text recognition; Vocabulary;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 1991., IEEE International Sympoisum on

Print_ISBN

0-7803-0050-5

Type

conf

DOI

10.1109/ISCAS.1991.176401

Filename

176401