Title
Initializing Subspace Constrained Gaussian Mixture Models
Abstract
A recent series of papers [1, 2, 3, 4] introduced Subspace Constrained Gaussian Mixture Models (SCGMMs) and showed that SCGMMs can very efficiently approximate Full Covariance Gaussian Mixture Models (FCGMMs); a significant reduction in the number of parameters is achieved with little loss in the accuracy of the model. SCGMMs were arrived at as a sequence of generalizations of diagonal covariance GMMs. As an artifact of this process the initialization of SCGMM parameters in that work is complex i.e., relies on best parameter settings of less general models. This paper overcomes this problem by showing how an FCGMM can be used to give a simple and direct initialization of an SCGMM. The initialization scheme is powerful enough that as the Dumber of parameters in an SCGMM approaches that of an FCGMM (i.e., large SCGMMs) further training of the SCGMM is unnecessary.
Year
DOI
Venue
2005
10.1109/ICASSP.2005.1415200
ICASSP '05). IEEE International Conference
Keywords
Field
DocType
Gaussian distribution,covariance matrices,parameter estimation,speech recognition,SCGMM,diagonal covariance GMM,initialization,parameter reduction,speech recognition,subspace constrained Gaussian mixture models
Subspace topology,Pattern recognition,Computer science,Probability distribution,Artificial intelligence,Estimation theory,Initialization,Covariance matrix,Hidden Markov model,Mixture model,Covariance
Conference
Volume
ISSN
ISBN
1
1520-6149
0-7803-8874-7
Citations 
PageRank 
References 
0
0.34
3
Authors
3
Name
Order
Citations
PageRank
Peder A. Olsen139837.80
Karthik Visweswariah240038.22
Ramesh Gopinath312310.65