Eigentriphone Modeling in Automatic Speech Recognition

PhD Thesis Proposal Defence


Title: "Eigentriphone Modeling in Automatic Speech Recognition"

by

Mr. Yu Ting KO


ABSTRACT:

In triphone-based acoustic modeling, it is difficult to robustly model 
infrequent triphones due to their lack of training samples. Naive 
maximum-likelihood (ML) estimation of infrequent triphone models produces 
poor triphone models and eventually affects the overall performance of an 
automatic speech recognition (ASR) system. Among different techniques 
proposed to solve the infrequent triphone problem, the most widely used 
method in current ASR systems is state tying because of its effectiveness 
in reducing model size and achieving good recognition results. However, 
state tying inevitably introduces quantization errors since triphones tied 
to the same state are not distinguishable in that state. On the other 
hand, speaker adaptation techniques have been well developed over the past 
decades. Speaker adaptation aims at adapting acoustic models to the 
characteristics of a particular speaker with a limited amount of speaker 
specific data. Motivated by the idea of these speaker adaptation 
techniques, we would like to solve the infrequent triphone problem from an 
adaptation point of view. In this proposal, we propose a new 
context-dependent modeling method called eigentriphones modeling. In 
contrast to the state tying method, all the triphone models are distinct 
from each other and thus may be more discriminative. The rational behind 
our method is that a basis is derived over the frequent triphones and each 
infrequent triphone is modeled as a point in the space spanned by the 
basis. The eigenvectors in the basis represent the most important 
context-dependent characteristics among the triphones and thus the 
infrequent triphones can be robustly modeled with few training samples. 
The proposed eigentriphone modeling was empirically evaluated on the Wall 
Street Journal 5K task and the TIMIT phoneme recognition task. It is shown 
that our proposed method consistently performs better than the most common 
state tying method. Future works for completion of the thesis are also 
given in the proposal.


Date:                   Monday, 13 May 2013

Time:                   10:00am - 12:00noon

Venue:                  Room 3405
                         lifts 17/18

Committee Members:      Dr. Brian Mak (Supervisor)
                         Prof. Dit-Yan Yeung (Chairperson)
 			Prof. Siu-Wing Cheng
 			Dr. Raymond Wong


**** ALL are Welcome ****