What question did this study set out to answer?

September 27, 1989

Corrective and reinforcement learning for speaker-independent continuous speech recognition

Key Points

This research aims to enhance speaker-independent continuous speech recognition by improving model training methods.
Extended the corrective training algorithm to speaker-independent continuous speech recognition.
Employed cross-validation to enhance the effective training size.
Introduced a near-miss sentence hypothesization algorithm for improving training accuracy.
Achieved over 20% error reduction in recognition accuracy with the proposed methods.
Reported improvements were consistent with and without the incorporation of grammar.

Abstract

Abstract This paper addresses the issue of learning hidden Markov model (HMM) parameters for speaker-independent continuous speech recognition. Bahl et al. (IEEE conference on acoustics, speech and signal processing, April 1988a) 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 continjous 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.

Mark Helpful

Bookmark

Relay