Paper Info
Title
Phoneme HMMs constrained by frame correlations
Abstract
Phoneme HMMs (hidden Markov models) that use correlations between two frames are proposed. The proposed technique constrains the output probability distributions of speaker-independent HMMs so that they are suitable for the input speaker. The speaker-dependent BC (bigram-constrained)-HMMs and speaker-independent BC-HMMs are generated from the conventional speaker-independent HMMs by combining the VQ (vector quantization)-code bigram (discrete case and tied-mixture case) or the conditional Gaussian density function (continuous case). The new models were evaluated by 23-phoneme recognition in continuous speech. In the speaker-dependent BC-HMMs, which use the speaker-dependent bigram created by 50 additional sentences of the test speaker, the best recognition accuracy of 74.8% was obtained by the tied-mixture type BC-HMMs. In the speaker-independent BC-HMMs, the best recognition accuracy of 67.5% was obtained by the continuous type BC-HMMs.<>
Year
DOI
Venue
1993
10.1109/ICASSP.1993.319274
Acoustics, Speech, and Signal Processing, 1993. ICASSP-93., 1993 IEEE International Conference  
Keywords
Field
DocType
constraint handling,correlation methods,hidden Markov models,speech recognition,vector quantisation,conditional Gaussian density function,correlations between two frames,hidden Markov models,output probability distributions,recognition accuracy,recognition in continuous speech,vector quantization
Training set,Pattern recognition,Computer science,Speech recognition,Gaussian,Vector quantization,Probability distribution,Bigram,Artificial intelligence,Estimation theory,Hidden Markov model,Probability density function
Conference
Volume
ISSN
ISBN
2
1520-6149
0-7803-0946-4
Citations 
PageRank 
References 
17
3.38
2
Authors
4
Name
Order
Citations
PageRank
Satoshi Takahashi1173.38
Tatsuo Matsuoka29120.76
Yasuhiro Minami311214.29
Kiyohiro Shikano42662928.81