DocumentCode :
2178155
Title :
Structured precision modelling with Cholesky Basis Superposition for speech recognition
Author :
Jia, Lei ; Yu, Kai ; Xu, Bo
Author_Institution :
Digital Media Content Technol. Center, C.A.S., Beijing, China
fYear :
2011
fDate :
22-27 May 2011
Firstpage :
5168
Lastpage :
5171
Abstract :
Structured precision modelling is an important approach to improve the intra-frame correlation modelling of the standard HMM, where Gaussian mixture model with diagonal covariance are used. Previous work has all been focused on direct structured representation of the precision matrices. In this paper, a new framework is pro posed, where the structure of the Cholesky square root of the precision matrix is investigated, referred to as Cholesky Basis Super position (CBS). Each Cholesky matrix associated with a particular Gaussian distribution is represented as a linear combination of a set of Gaussian independent basis upper-triangular matrices. Efficient optimization methods are derived for both combination weights and basis matrices. Experiments on a Chinese dictation task showed that the proposed approach can significantly outperformed the direct structured precision modelling with similar number of parameters as well as full covariance modelling.
Keywords :
Gaussian distribution; matrix algebra; optimisation; speech recognition; CBS; Cholesky basis superposition; Cholesky matrix; Cholesky square; Gaussian distribution; Gaussian independent basis upper-triangular matrix; Gaussian mixture model; HMM; direct structured precision modelling; optimization methods; precision matrix; speech recognition; structured precision modelling; Covariance matrix; Equations; Hidden Markov models; Mathematical model; Optimization; Symmetric matrices; Unsolicited electronic mail; Cholesky square root; inverse covariance modeling; precision modelling;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Acoustics, Speech and Signal Processing (ICASSP), 2011 IEEE International Conference on
Conference_Location :
Prague
ISSN :
1520-6149
Print_ISBN :
978-1-4577-0538-0
Electronic_ISBN :
1520-6149
Type :
conf
DOI :
10.1109/ICASSP.2011.5947521
Filename :
5947521
Link To Document :
بازگشت