A number of decoding strategies for large vocabulary speech recognition are examined from the viewpoint of their search space representation. Different design solutions are compared with respect to the integration of linguistic and acoustic constraints, as implied by M-gram LMs and cross-word phonetic contexts. This study is articulated along two main axes, namely, the network expansion and the search algorithm itself. Three broad classes of decoding methods are reviewed: the use of weighted finite state transducers for static network expansion, the time-synchronous dynamic-expansion search and the asynchronous stack decoding.
Cite as: Aubert, X.L. (2000) A brief overview of decoding techniques for large vocabulary continuous speech recognition. Proc. ASR2000 - Automatic Speech Recognition: Challenges for the New Millenium, 91-96
@inproceedings{aubert00_asr,
author={Xavier L. Aubert},
title={{A brief overview of decoding techniques for large vocabulary continuous speech recognition}},
year=2000,
booktitle={Proc. ASR2000 - Automatic Speech Recognition: Challenges for the New Millenium},
pages={91--96}
}