Speech Recognition Using SVMs
Abstract
An important issue in applying SVMs to speech recognition is the ability to classify variable length sequences. This paper presents extensions to a standard scheme for handling this variable length data, the Fisher score. A more useful mapping is introduced based on the likelihood-ratio. The score-space defined by this mapping avoids some limitations of the Fisher score. Class-conditional gen(cid:173) erative models are directly incorporated into the definition of the score-space. The mapping, and appropriate normalisation schemes, are evaluated on a speaker-independent isolated letter task where the new mapping outperforms both the Fisher score and HMMs trained to maximise likelihood.
Cite
Text
Smith and Gales. "Speech Recognition Using SVMs." Neural Information Processing Systems, 2001.Markdown
[Smith and Gales. "Speech Recognition Using SVMs." Neural Information Processing Systems, 2001.](https://mlanthology.org/neurips/2001/smith2001neurips-speech/)BibTeX
@inproceedings{smith2001neurips-speech,
title = {{Speech Recognition Using SVMs}},
author = {Smith, N. and Gales, Mark},
booktitle = {Neural Information Processing Systems},
year = {2001},
pages = {1197-1204},
url = {https://mlanthology.org/neurips/2001/smith2001neurips-speech/}
}