This paper addresses the issue of learning hidden Markov model (HMM) parameters for speaker-independent continuous speech recognition. Bahl et al. [1] introduced the corrective training algorithm for speaker-dependent isolated word recognition. Their algorithm attempted to improve the recognition accuracy on the training data. In this work, we extend this algorithm to speaker-independent continuous speech recognition. We use cross-validation to increase the effective training size. We also introduce a near-miss sentence hypothesization algorithm for continuous speech training. The combination of these two approaches resulted in over 20% error reductions both with and without grammar.
Cite as: Lee, K.-F., Mahajan, S. (1989) Corrective and reinforcement learning for speaker-independent continuous speech recognition. Proc. First European Conference on Speech Communication and Technology (Eurospeech 1989), 1490-1493, doi: 10.21437/Eurospeech.1989-57
@inproceedings{lee89b_eurospeech,
author={Kai-Fu Lee and Sanjoy Mahajan},
title={{Corrective and reinforcement learning for speaker-independent continuous speech recognition}},
year=1989,
booktitle={Proc. First European Conference on Speech Communication and Technology (Eurospeech 1989)},
pages={1490--1493},
doi={10.21437/Eurospeech.1989-57}
}