DocumentCode
2686757
Title
A real-time Chinese speech recognition IC with double mixtures
Author
Wei Han ; Cheong-Fat Chan ; Tan Lee ; Chiu-Sing Choy ; Kong-Pang Pun ; Ching, P.C.
Author_Institution
Dept. of Electron. Eng., Chinese Univ. of Hong Kong, China
Volume
2
fYear
2003
fDate
21-24 Oct. 2003
Firstpage
926
Abstract
A hidden Markov model (HMM) based isolated word recognizer IC is presented in this paper. The test chip was fabricated with a 0.35 μm CMOS technology. In this design double-mixture probability density and table look-up method are employed. The chip can operate at 20MHz at 3.3V and at this frequency the recognition time is 0.5 sec for a 50-word speech library. Tested with 353 speech data from AURORA 2 databases, the chip´s recognition accuracy is 93.8%. This new IC can be used in Cantonese speech recognition.
Keywords
CMOS integrated circuits; hidden Markov models; speech recognition; table lookup; 0.35 microns; 0.5 sec; 20 MHz; 3.3 V; 93.8 percent; AURORA 2 databases; CMOS technology; Cantonese speech recognition; Chinese speech recognition IC; HMM; double mixtures; double-mixture probability density; hidden Markov model; isolated word recognizer IC; speech library; table look-up method; test chip;
fLanguage
English
Publisher
ieee
Conference_Titel
ASIC, 2003. Proceedings. 5th International Conference on
ISSN
1523-553X
Print_ISBN
0-7803-7889-X
Type
conf
DOI
10.1109/ICASIC.2003.1277362
Filename
1277362
Link To Document