INTERSPEECH
2010
12 years 11 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...