SuperLectures.com

DEFINING THE CONTROLLING PARAMETER IN CONSTRAINED DISCRIMINATIVE LINEAR TRANSFORM FOR SUPERVISED SPEAKER ADAPTATION

Full Paper at IEEE Xplore

Adaptation for ASR

Presented by: Dimitri Kanevsky, Author(s): Danning Jiang, IBM China Research Lab, China; Dimitri Kanevsky, Emmanuel Yashchin, IBM T.J. Watson Research Center, United States; Yong Qin, IBM China Research Lab, China

Constrained discriminative linear transform (CDLT) optimized with Extended Baum-Welch (EBW) has been presented in the literature as a discriminative speaker adaptation method that outperforms the conventional maximum likelihood algorithm. Defining the controlling parameter of EBW to achieve the best performance of speaker adaptation, however, still remains an open question. This paper presents an empirical study on this issue. Results of our experiment suggest that a log-linear relationship exists between the optimal controlling parameter and the amount of data. This relationship can be used to efficiently define the controlling parameter for each test speaker to improve CDLT performance. We also discuss the possibility of generalizing the log-linear rule to a wider range of learning problems because such knowledge can substantially reduce the computation effort for parameter tuning.


  Speech Transcript

|

  Comments

Please sign in to post your comment!

  Lecture Information

Recorded: 2011-05-24 16:35 - 16:55, Panorama
Added: 15. 6. 2011 14:45
Number of views: 80
Video resolution: 1024x576 px, 512x288 px
Video length: 0:16:01
Audio track: MP3 [5.39 MB], 0:16:01