Presentation 2005/12/15
Evaluation of Front-End Based on Speech Feature Extraction Method SPADE Using Standard Noise Robustness Evaluation Corpus
Kentaro ISHIZUKA, Tomohiro NAKATANI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper proposes noise-robust front-end processing combining speech feature extraction method 'SPADE' with noise suppression and feature normalization methods. A nonlinear spectral subtraction or an adaptive Wiener filter method was adopted as noise suppression. The cepstral mean or variance normalization method was adopted as feature normalization methods. As results from the evaluation using AURORA-2J, the proposed method achieves word accuracies of 82.58% for clean training condition, and 92.55% for multicondition training condition, at highest.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Speech feature / Noise robust frontend / Subband / Periodicity / Aperiodicity
Paper # NLC2005-89,SP2005-122
Date of Issue

Conference Information
Committee NLC
Conference Date 2005/12/15(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 Natural Language Understanding and Models of Communication (NLC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Evaluation of Front-End Based on Speech Feature Extraction Method SPADE Using Standard Noise Robustness Evaluation Corpus
Sub Title (in English)
Keyword(1) Speech feature
Keyword(2) Noise robust frontend
Keyword(3) Subband
Keyword(4) Periodicity
Keyword(5) Aperiodicity
1st Author's Name Kentaro ISHIZUKA
1st Author's Affiliation NTT Communication Science Laboratories, NTT Corporation()
2nd Author's Name Tomohiro NAKATANI
2nd Author's Affiliation NTT Communication Science Laboratories, NTT Corporation
Date 2005/12/15
Paper # NLC2005-89,SP2005-122
Volume (vol) vol.105
Number (no) 494
Page pp.pp.-
#Pages 2
Date of Issue