Power exponential densities for the training and classification of acoustic feature vectors in speech recognition

Abstract

We consider a parametric family of multivariate density functions formed by mixture models from univariate functions of the type exp(–|x|α) for modeling acoustic feature vectors used in automatic recognition of speech. The parameter α is used to measure the non-Gaussian nature of the data. Previous work has focused on estimating the mean and the variance of the data for a fixed α. Here we attempt to estimate the α from the data using a maximum likelihood criterion. Among other things, we show that there is a balance between α and the number of data pointsNthat must be satisfied for efficient estimation. Numerical experiments are performed on multidimensional vectors obtained from speech data. © 2001 American Statistical Association, Institute of Mathematical Statistics, and Interface Foundation of North America.