In this paper we present an approach that makes use of both Bayesian predictive classification (BPC) and parallel model combination (PMC) to achieve increased robustness towards noise. PMC provides a method for finding parameter estimates for speech corrupted by noise, while BPC is a method that compensates for uncertainty of parameter estimates. Thus, these methods can be combined in order to obtain knowledge about the mismatch situation and simultaneously account for uncertainty in this knowledge. We apply this technique in an unsupervised approach on the Aurora2 database and show that good performance is obtained.
Cite as: Pettersen, S.G., Johnsen, M.H., Myrvoll, T.A. (2005) Joint Bayesian predictive classification and parallel model combination for robust speech recognition. Proc. Interspeech 2005, 373-376, doi: 10.21437/Interspeech.2005-211
@inproceedings{pettersen05_interspeech,
author={Svein G. Pettersen and Magne H. Johnsen and Tor A. Myrvoll},
title={{Joint Bayesian predictive classification and parallel model combination for robust speech recognition}},
year=2005,
booktitle={Proc. Interspeech 2005},
pages={373--376},
doi={10.21437/Interspeech.2005-211}
}