Invited: Robust Acoustic Modeling for Speech Recognition (国際ワークショップ"Beyond HMM") Robust Acoustic Modeling for Speech Recognition

Access this Article

Search this Article

Author(s)

Abstract

While Hidden Markov Models (HMMs) have been successfully applied to automatic speech recognition they are not still robust enough against differences in speaking-styles and environmental noises. To tackle this problem we need to study the inner structure of speech by using large corpus and rich computational power. In this direction the model size tends to be increase and hence the data insufficiency problem becomes more serious. In this paper we focus on robust modeling against data insufficiency. Approaches based on information criteria such as Minimum Description Length and structural apporoaches in which models are changed according to the amount of data availabl are discussed. While these techniques have been important for HMM research it will be more important in the research beyond HMM.While Hidden Markov Models (HMMs) have been successfully applied to automatic speech recognition, they are not still robust enough against differences in speaking-styles, and environmental noises. To tackle this problem, we need to study the inner structure of speech by using large corpus and rich computational power. In this direction, the model size tends to be increase and hence the data insufficiency problem becomes more serious. In this paper, we focus on robust modeling against data insufficiency. Approaches based on information criteria such as Minimum Description Length and structural apporoaches in which models are changed according to the amount of data availabl are discussed. While these techniques have been important for HMM research, it will be more important in the research beyond HMM.

While Hidden Markov Models (HMMs) have been successfully applied to automatic speech recognition, they are not still robust enough against differences in speakers, speaking-styles, and environmental noises. To tackle this problem, we need to study the inner structure of speech by using large corpus and rich computational power. In this direction, the model size tends to be increase and hence the data insufficiency problem becomes more serious. In this paper, we focus on robust modeling against data insufficiency. Approaches based on information criteria such as Minimum Description Length and structural approaches in which models are changed according to the amount of data availabl are discussed. While these techniques have been important for HMM research, it will be more important in the research beyond HMM.

Journal

  • IPSJ SIG Notes

    IPSJ SIG Notes 2004(131(2004-SLP-054)), 7-12, 2004-12-20

    Information Processing Society of Japan (IPSJ)

References:  22

Codes

  • NII Article ID (NAID)
    110002950575
  • NII NACSIS-CAT ID (NCID)
    AN10442647
  • Text Lang
    ENG
  • Article Type
    Technical Report
  • ISSN
    09196072
  • NDL Article ID
    7213338
  • NDL Source Classification
    ZM13(科学技術--科学技術一般--データ処理・計算機)
  • NDL Call No.
    Z14-1121
  • Data Source
    CJP  NDL  NII-ELS  IPSJ 
Page Top