We address a method to efficiently select Gaussian mixtures for fast acoustic likelihood computation. It makes use of context-independent models for selection and back-off of corresponding triphone models. Specifically, for the k-best phone models by the preliminary evaluation, triphone models of higher resolution are applied, and others are assigned likelihoods with the monophone models. This selection scheme assigns more reliable back-off likelihoods to the un-selected states than the conventional Gaussian selection based on a VQ codebook. Experimental results show that this method can achieves a comparable performance, and works much better under the aggressive pruning condition. Together with the phonetic tied-mixture (PTM) modeling, acoustic matching cost is reduced to almost 14 with little loss of accuracy.
展开▼
