Presentation	1997/6/20 A RESTRUCTURING OF GAUSSIAN MIXTURE PDFS IN SPEAKER-INDEPENDENT ACOUSTIC MODELS Atsushi Nakamura,
PDF Download Page	PDF download Page Link
Abstract(in Japanese)	(See Japanese page)
Abstract(in English)	In continuous speech recognition featuring hidden Markov model (HMM), word N-gram and time-synchronous beam search, a local modeling mismatch in the HMM often causes the recognition performance to degrade. To cope with such local modeling mismatches, this paper proposes a method of restructuring Gaussian mixture pdfs in a speaker-independent HMM based on speech samples. In this method, mixture components are copied and shared among multiple mixture pdfs, with taking into account the tendency of local errors given by comparing a pre-trained HMM and speech samples. Experimental results have proven that the proposed method can effectively restore local modeling mismatches and improve the recognition performance.
Keyword(in Japanese)	(See Japanese page)
Keyword(in English)	speaker-independent speech recognition / spontaneous speech / HMM / recognition error tendency
Paper #	SP97-19
Date of Issue

Conference Information
Committee	SP
Conference Date	1997/6/20(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)	A RESTRUCTURING OF GAUSSIAN MIXTURE PDFS IN SPEAKER-INDEPENDENT ACOUSTIC MODELS
Sub Title (in English)
Keyword(1)	speaker-independent speech recognition
Keyword(2)	spontaneous speech
Keyword(3)	HMM
Keyword(4)	recognition error tendency
1st Author's Name	Atsushi Nakamura
1st Author's Affiliation	ATR Interpreting Telecommunications Research Laboratories()
Date	1997/6/20
Paper #	SP97-19
Volume (vol)	vol.97
Number (no)	115
Page	pp.pp.-
#Pages	8
Date of Issue