Presentation 1997/1/16
High-speed Segment Quantization Based on KL-expansion and Generalized Probabilistic Descent Method
Tsuneo NITTA, Akinori KAWAMURA, Yasuyuki MASAI, Akira NAKAYAMA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The hybrid algorithm of SMQ (Statistical Matrix Quantization) and HMM shows high performance in vocabulary-unspecific, speaker-independent speech recognition, however, it needs lots of computation and memory at a segment quantizer of SMQ. In this paper, we propose a newly developed, two-stage segment quantizer with a feature extractor based on KL-expansion and a classifier, both are traind by using competitive training of MCE/GPD. Result of experiments shows 1/30 1/40 reduction in both computation time and a memory size with the same perfomance that the old version of SMQ shows.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Speech Recognition / HMM / Segment Quantization / Competitive Training / KL-expansion / GPD
Paper # SP96-94
Date of Issue

Conference Information
Committee SP
Conference Date 1997/1/16(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Speech (SP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) High-speed Segment Quantization Based on KL-expansion and Generalized Probabilistic Descent Method
Sub Title (in English)
Keyword(1) Speech Recognition
Keyword(2) HMM
Keyword(3) Segment Quantization
Keyword(4) Competitive Training
Keyword(5) KL-expansion
Keyword(6) GPD
1st Author's Name Tsuneo NITTA
1st Author's Affiliation Toshiba Multimedia Eng. Lab.()
2nd Author's Name Akinori KAWAMURA
2nd Author's Affiliation Toshiba Multimedia Eng. Lab.
3rd Author's Name Yasuyuki MASAI
3rd Author's Affiliation Toshiba Multimedia Eng. Lab.
4th Author's Name Akira NAKAYAMA
4th Author's Affiliation Toshiba Multimedia Eng. Lab.
Date 1997/1/16
Paper # SP96-94
Volume (vol) vol.96
Number (no) 448
Page pp.pp.-
#Pages 8
Date of Issue