INTERSPEECH
2010
13 years 2 months ago
2010
In this paper, we propose a new semi-supervised training method for Gaussian Mixture Models. We add a conditional entropy minimizer to the maximum mutual information criteria, whi...