Deterministic Annealing EM Algorithm in Acoustic Modeling for Speaker and Speech Recognition

  • Authors:

  • Yohei Itaya;Heiga Zen;Yoshihiko Nankaku;Chiyomi Miyajima;Keiichi Tokuda;Tadashi Kitamura

  • Affiliations:

  • The authors are with the Department of Computer Science and Engineering, Nagoya Institute of Technology, Nagoya-shi, 466-8555 Japan. E-mail: yoheir32@ics.nitech.ac.jp, E-mail: zen@ics.nitech.ac.jp ...;The authors are with the Department of Computer Science and Engineering, Nagoya Institute of Technology, Nagoya-shi, 466-8555 Japan. E-mail: yoheir32@ics.nitech.ac.jp, E-mail: zen@ics.nitech.ac.jp ...;The authors are with the Department of Computer Science and Engineering, Nagoya Institute of Technology, Nagoya-shi, 466-8555 Japan. E-mail: yoheir32@ics.nitech.ac.jp, E-mail: zen@ics.nitech.ac.jp ...;The author is with the Department of Media Science, Nagoya University, Nagoya-shi, 466-8603 Japan. E-mail: miyajima@is.nagoya-u.ac.jp;The authors are with the Department of Computer Science and Engineering, Nagoya Institute of Technology, Nagoya-shi, 466-8555 Japan. E-mail: yoheir32@ics.nitech.ac.jp, E-mail: zen@ics.nitech.ac.jp ...;The authors are with the Department of Computer Science and Engineering, Nagoya Institute of Technology, Nagoya-shi, 466-8555 Japan. E-mail: yoheir32@ics.nitech.ac.jp, E-mail: zen@ics.nitech.ac.jp ...

  • Venue:
  • IEICE - Transactions on Information and Systems
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper investigates the effectiveness of the DAEM (Deterministic Annealing EM) algorithm in acoustic modeling for speaker and speech recognition. Although the EM algorithm has been widely used to approximate the ML estimates, it has the problem of initialization dependence. To relax this problem, the DAEM algorithm has been proposed and confirmed the effectiveness in artificial small tasks. In this paper, we applied the DAEM algorithm to practical speech recognition tasks: speaker recognition based on GMMs and continuous speech recognition based on HMMs. Experimental results show that the DAEM algorithm can improve the recognition performance as compared to the standard EM algorithm with conventional initialization algorithms, especially in the flat start training for continuous speech recognition.