Maximum Likelihood Training of Bases for Rapid Adaptation

@inproceedings{Visweswariah2002MaximumLT,
  title={Maximum Likelihood Training of Bases for Rapid Adaptation},
  author={Karthik Visweswariah and V. Goel and Ramesh A. Gopinath},
  year={2002}
}
Adaptation using linear transforms is well known to significantly improve the performance of speech recognition systems. Restricting the transforms to lie in a subspace obtained at training time has been shown to be useful in achieving good performance even with few seconds of adaptation data. This paper addresses the issue of how the subspace is learned. Algorithms for maximum likelihood (ML) estimation of the basis (which defines the subspace) are given. Our experimental results suggest that… CONTINUE READING

From This Paper

Figures, tables, and topics from this paper.

Similar Papers

Loading similar papers…