PaperPower exponential densities for the training and classification of acoustic feature vectors in speech recognition
PaperPenalized maximum-likelihood estimation, the Baum-Welch algorithm, diagonal balancing of symmetric matrices and applications to training acoustic data