可変音声分布を用いたMAP推定に基づく音声強調(スマートパーソナルシステム,一般)  [in Japanese] Speech Enhancement by MAP Estimation Using a Variable Speech Distribution  [in Japanese]

Access this Article

  • CiNii Fulltext PDF

    Subscription

Search this Article

Abstract

本稿ではMAP推定に基づく新たな音声強調手法を提案する.従来は長時間音声データによって音声スペクトル分布を一意に決定していたため,音声区間における音質や非音声区間における雑音除去性能が劣化するという問題があった.この問題を解決するために,周波数全体のパワースペクトルを基準とし音声スペクトル分布を動的に変化させる方法を提案する.提案法では,音声スペクトル分布を非音声区間では指数分布に近づけることによって,優れた雑音除去性能を実現する.また,音声区間ではレーリー分布に近づけることによって,音質の劣化を抑えることが可能となる.

A novel speech enhancement algorithm by MAP estimation is proposed in this paper. In the previous method, the performance of noise reduction in non-speech segment or the quality of the output speech in speech segment depends on the probability density function (PDF) of the speech spectral amplitude which is calculated and fixed beforehand. To realize both the high speech quality and noise reduction performance at the same time, we introduce a variable speech spectral distribution adjusted by the sum of power spectral densities. In a non-speech segment, the PDF of the speech spectral amplitude approaches to an exponential density to storongly reduce noise the noise. While in a speech segment, it approaches to a Rayleigh density to improve the quality of the enhanced speech.

Journal

IEICE technical report   [List of Volumes]

IEICE technical report 106(95), 43-48, 2006-06-08  [Table of Contents]

The Institute of Electronics, Information and Communication Engineers

References:  6

You must have a user ID to see the references.If you already have a user ID, please click "Login" to access the info.New users can click "Sign Up" to register for an user ID.

Codes

  • NII Article ID (NAID) :
    110004751118
  • NII NACSIS-CAT ID (NCID) :
    AA1196239X
  • Text Lang :
    JPN
  • Article Type :
    ART
  • ISSN :
    09135685
  • NDL Article ID :
    7976145
  • NDL Source Classification :
    ZN33(科学技術--電気工学・電気機械工業--電子工学・電気通信)
  • NDL Call No. :
    Z16-940
  • Databases :
    CJP  NDL  NII-ELS