A FAST, ACCURATE APPROXIMATION TO LOG LIKELIHOOD OF GAUSSIAN MIXTURE MODELS

It has been a common practice in speech recognition and elsewhere to approximate the log likelihood of a Gaussian mixture model (GMM) with the maximum component log likelihood. While often a computational necessity, the max approximation comes at a price of inferior modeling when the Gaussian components significantly over-lap. This paper shows how the approximation error can be reduced by changing component priors. In our experiments the loss in word error rate due to max approximation, albeit small, is reduced by 50-100 % at no cost in computational efficiency. Furthermore, we expect acoustic models will become larger with time and increase compo-nent overlap and word error rate loss. This makes reducing the ap-proximation error more relevant. The techniques considered do not use the original data and can easily be applied as a post-processing step to any GMM. Index Terms — Gaussian mixture model, acoustic model, maximum approximation, exponential distribution 1