Presentation 2004/6/18
Running Spectrum Analysis for Robust Speech Recognition
Kuniko ISHIDA, Noboru HAYASAKA, Naoya Yoshikazu, Yoshikazu MIYANAGA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this report, a noise robust speech recognition system which has been already developed is extended. The noise robust speech mechanism in the recognition system has been designed on the moderation spectrum domain. By using high order FIR filtering method (RSF), any time varying noise components can be eliminated and by using dynamic rang adjustment method (DRA). Any influences on noise SNR can be compensated. However, these methods require quite high calculation cost, i.e., a large number of coefficients in RSF. In this report, the novel approach in which low calculation cost is only required, i.e., Running Spectrum Analysis (RSA) which employs only FFT in its processing, is developed. By using this proposed algorithm, the similar performance can be realized.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Robust speech recognition / MFCC / RSF / Modulation spectrum
Paper # SIS2004-5
Date of Issue

Conference Information
Committee SIS
Conference Date 2004/6/18(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 Smart Info-Media Systems (SIS)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Running Spectrum Analysis for Robust Speech Recognition
Sub Title (in English)
Keyword(1) Robust speech recognition
Keyword(2) MFCC
Keyword(3) RSF
Keyword(4) Modulation spectrum
1st Author's Name Kuniko ISHIDA
1st Author's Affiliation Graduate School of Information and Science Technology, Hokkaido University()
2nd Author's Name Noboru HAYASAKA
2nd Author's Affiliation Graduate School of Information and Science Technology, Hokkaido University
3rd Author's Name Naoya Yoshikazu
3rd Author's Affiliation Graduate School of Information and Science Technology, Hokkaido University
4th Author's Name Yoshikazu MIYANAGA
4th Author's Affiliation Graduate School of Information and Science Technology, Hokkaido University
Date 2004/6/18
Paper # SIS2004-5
Volume (vol) vol.104
Number (no) 144
Page pp.pp.-
#Pages 6
Date of Issue