Speech Feature Analysis Using Variational Bayesian PCA

In most hidden Markov based automatic speech recognition systems one of the fundamental question is to determine the intrinsic speech feature dimensionality and the number of clusters used in the Gaussian mixture model. We analyzed mel-frequency band energies using a variational Bayesian principal component analysis method to estimate the feature dimensionality as well as the number of Gaussian mixtures by learning a maximum lower bound of the evidence instead of maximizing the likelihood function as used in conventional speech recognition systems. In analyzing the TIMIT speech data set, our method revealed the intrinsic structures of vowels and consonants. The usefulness of this method is demonstrated in the superior classification performance for recognition the most difficult phonemes /b/, /d/ and /g/.