Hidden Markov Models for Speech Recognition
Hidden Markov Models for Speech Recognition
The acoustic-modeling problem in automatic speech recognition
The acoustic-modeling problem in automatic speech recognition
The Lincoln tied-mixture HMM continuous speech recognizer
ICASSP '91 Proceedings of the Acoustics, Speech, and Signal Processing, 1991. ICASSP-91., 1991 International Conference
Phonemic HMM constrained by statistical VQ-code transition
ICASSP'92 Proceedings of the 1992 IEEE international conference on Acoustics, speech and signal processing - Volume 1
| Hi-index | 0.00 |
This paper proposes new Hidden Markov Models (HMMs) that use correlations between two frames to constrain the feature distributions to the region that is appropriate for an input speaker. This makes it possible to reduce the overlapping of feature distributions between different phonemes. In ICASSP92, we proposed the bigram-constrained HMM based on the combination of the discrete speaker-independent HMM and the VQ-code bigram, and showed that it performed better than a conventional speaker-independent HMM. In this paper, tied-mixture HMMs are adopted to create the tied-mixture type bigram-constrained HMM to obtain better recognition performance. Furthermore, the strategy is extended to the continuous HMM. These three types of HMMs are formulated and evaluated by phoneme recognition in continuous speech.