Paper Info
Title
Model Adaptation for Long Convolutional Distortion by Maximum Likelihood Based State Filtering Approach
Abstract
In environment with considerably long reverberation time, each frame of speech is affected by energy components from the preceding frames. Therefore, to adapt parameters of a state of HMM, it becomes necessary to consider these frames, and compute their contributions to current state. However, these speech frames preceding to a state of HMM are not known during adaptation of the models. In this paper, we propose to use preceding states as units of preceding speech segments, estimate their contributions to current state in maximum like- lihood manner, and adapt models by accounting their con- tributions. When clean models were adapted by proposed method for a speaker-dependent isolated word recognition task, word accuracy of the system typically increased from 67.6% to 83.2%, and from 44.8% to 72.5%, for channel distorted speech simulated by linear convolution of clean speech and impulse responses with reverberation time (T60) of 310 ms and 780 ms, respectively.
Year
DOI
Venue
2006
10.1109/ICASSP.2006.1660225
Acoustics, Speech and Signal Processing, 2006. ICASSP 2006 Proceedings. 2006 IEEE International Conference
Keywords
Field
DocType
filtering theory,hidden Markov models,maximum likelihood estimation,speech processing,speech recognition,HMM,channel distorted speech,long convolutional distortion,maximum likelihood approach,preceding speech segments,reverberation time,speaker-dependent isolated word recognition,state filtering approach
Speech processing,Reverberation,Pattern recognition,Computer science,Convolution,Word recognition,Filter (signal processing),Impulse (physics),Speech recognition,Artificial intelligence,Hidden Markov model,Distortion
Conference
Volume
ISSN
ISBN
1
1520-6149
1-4244-0469-X
Citations 
PageRank 
References 
20
1.12
9
Authors
3
Name
Order
Citations
PageRank
Chandra Kant Raut1201.12
Takuya Nishimoto222728.95
Shigeki Sagayama3201.12