機率式調變頻譜分解於強健性語音辨識 Probabilistic Modulation Spectrum Factorization for Robust Speech Recognition

@inproceedings{2011PM,
  title={機率式調變頻譜分解於強健性語音辨識 Probabilistic Modulation Spectrum Factorization for Robust Speech Recognition},
  author={朱紋儀 高予真 and 陳柏琳 國立臺灣師範大學資訊工程學系},
  year={2011}
}
  • 朱紋儀 高予真, 陳柏琳 國立臺灣師範大學資訊工程學系
  • Published 2011
摘要 在自動語音辨識技術的發展上,語音強健性一直都是一門重要的研究議題。在眾多的強 健性技術中,針對語音特徵參數進行強化與補償為其中之一大主要派別。其中,近年來 已有為數不少的新方法,藉由更新語音特徵時間序列及其調變頻譜來提升語音特徵的強 健性。本論文即是從語音特徵時間序列的調變頻譜域著手,採用機率式潛藏語意分析之 概念,對調變頻譜施以機率式分解並進行成分分析、進而擷取出較重要的成分以求得更 具強健性的語音特徵。本方法之所有實驗皆於國際通用的 Aurora-2 連續數字資料庫進 行,相較於使用梅爾倒頻譜特徵之基礎實驗,本方法能達到 62.84%的相對錯誤降低率。 此外,我們也嘗試將所提方法跟一些知名的特徵強健技術做結合;實驗顯示,相對於單 一方法而言,此結合法可進一步提升辨識精確率,代表所提之新方法與許多特徵強健技 術有良好的加成性。 關鍵詞: 雜訊強健性、語音特徵參數強化、調變頻譜、機率式潛藏語意分析 

From This Paper

Topics from this paper.

References

Publications referenced by this paper.
Showing 1-10 of 19 references

Segmental Feature Vector Normalization for Noise Robust Speech Recognition

A. Vikki, K. Laurila
Speech Communication, Vol. 25: pp. 133-147, 1998 • 1998
View 10 Excerpts
Highly Influenced

Normalization of the Speech Modulation Spectra for Robust Speech Recognition

IEEE Transactions on Audio, Speech, and Language Processing • 2008
View 6 Excerpts
Highly Influenced

Modulation spectrum equalization for robust speech recognition

2007 IEEE Workshop on Automatic Speech Recognition & Understanding (ASRU) • 2007
View 5 Excerpts
Highly Influenced

Cepstral Analysis Techniques for Automatic Speaker Verification

S. Furui
IEEE Trans. on Acoustic, Speech and Signal Processing, Vol. 29(2): pp. 254-272, 1981 • 1981
View 6 Excerpts
Highly Influenced

Word Topic Models for Spoken Document Retrieval and Transcription

ACM Trans. Asian Lang. Inf. Process. • 2009
View 3 Excerpts
Highly Influenced

Histogram equalization of speech representation for robust speech recognition

IEEE Transactions on Speech and Audio Processing • 2005
View 3 Excerpts
Highly Influenced

Similar Papers

Loading similar papers…