ISCA Archive Interspeech 2008
ISCA Archive Interspeech 2008

Monte Carlo model-space noise adaptation for speech recognition

Daniel Povey, Brian Kingsbury

We describe a Monte Carlo method formodel-space noise adaptation of Gaussian mixture models (GMMs). This method combines a single-Gaussian noise model with the GMM speech model to produce an adapted model. It is similar to Parallel Model Combination or model-space Joint, except that it applies to spliced and projected MFCC features rather than to MFCC plus dynamic features. We demonstrate the necessity of re-estimating the noise using both the silence and speech frames rather than just estimating it from silence frames, and obtain improvements on a matched test set without added noise using a system that includes all standard adaptation techniques.


doi: 10.21437/Interspeech.2008-308

Cite as: Povey, D., Kingsbury, B. (2008) Monte Carlo model-space noise adaptation for speech recognition. Proc. Interspeech 2008, 1281-1284, doi: 10.21437/Interspeech.2008-308

@inproceedings{povey08_interspeech,
  author={Daniel Povey and Brian Kingsbury},
  title={{Monte Carlo model-space noise adaptation for speech recognition}},
  year=2008,
  booktitle={Proc. Interspeech 2008},
  pages={1281--1284},
  doi={10.21437/Interspeech.2008-308}
}