Bayesian linear regression for hidden Markov model based on optimizing variational bounds

Shinji Watanabe*, Atsushi Nakamura, Biing Hwang Juang

*この研究の対応する著者

研究成果: Conference contribution

12 被引用数 (Scopus)

抄録

Linear regression for Hidden Markov Model (HMM) parameters is widely used for the adaptive training of time series pattern analysis especially for speech processing. This paper realizes a fully Bayesian treatment of linear regression for HMMs by using variational techniques. This paper analytically derives the variational lower bound of the marginalized log-likelihood of the linear regression. By using the variational lower bound as an objective function, we can optimize the model topology and hyper-parameters of the linear regression without controlling them as tuning parameters; thus, we realize linear regression for HMM parameters in a non-parametric Bayes manner. Experiments on large vocabulary continuous speech recognition confirm the generalizability of the proposed approach, especially for small quantities of adaptation data.

本文言語English
ホスト出版物のタイトル2011 IEEE International Workshop on Machine Learning for Signal Processing - Proceedings of MLSP 2011
DOI
出版ステータスPublished - 2011
外部発表はい
イベント21st IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2011 - Beijing, China
継続期間: 2011 9月 182011 9月 21

出版物シリーズ

名前IEEE International Workshop on Machine Learning for Signal Processing

Other

Other21st IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2011
国/地域China
CityBeijing
Period11/9/1811/9/21

ASJC Scopus subject areas

  • 人間とコンピュータの相互作用
  • 信号処理

フィンガープリント

「Bayesian linear regression for hidden Markov model based on optimizing variational bounds」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル