105
click to vote
EMNLP
2007
15 years 4 months ago
2007
We present a maximally streamlined approach to learning HMM-based acoustic models for automatic speech recognition. In our approach, an initial monophone HMM is iteratively refin...