This paper addresses the issues in recognition of subword units of speech using support vector machines. Discriminative training and good generalization capability of support vector machines are useful in developing acoustic models for subword units when the confusability among the units is high. We compare the performance of support vector machine based systems with that of hidden Markov models in recognition of subword units. We demonstrate the better performance of support vector machines in recognition of monophone units in a large corpus of Japanese speech and recognition of consonant-vowel units in a broadcast news corpus of an Indian language.
Cite as: Chandra Sekhar, C., Lee, W.F., Takeda, K., Itakura, F. (2003) Acoustic modeling of subword units usingsupport vector machines. Proc. Workshop on Spoken Language Processing, 79-86
@inproceedings{chandrasekhar03_wslp,
author={C. {Chandra Sekhar} and W. F. Lee and K. Takeda and Fumitada Itakura},
title={{Acoustic modeling of subword units usingsupport vector machines}},
year=2003,
booktitle={Proc. Workshop on Spoken Language Processing},
pages={79--86}
}