Posterior-based feature with partial distance elimination for speech recognition

摘要

A high-dimensional posterior-based feature with partial distance elimination may be utilized for speech recognition. The log likelihood values of a large number of Gaussians are needed to generate the high-dimensional posterior feature. Gaussians with very small log likelihoods are associated with zero posterior values. Log likelihoods for Gaussians for a speech frame may be evaluated with a partial distance elimination method. If the partial distance of a Gaussian is already too small, the Gaussian will have a zero posterior value. The partial distance may be calculated by sequentially adding individual dimensions in a group of dimensions. The partial distance elimination occurs when less than all of the dimensions in the group are sequentially added.

主权项

1. A method of utilizing a posterior-based feature with partial distance elimination for speech recognition, comprising:
receiving, by a computer, an utterance comprising a plurality of speech frames; and evaluating, by the computer, a plurality of log likelihoods of Gaussians for a speech frame to calculate a partial distance by sequentially adding a plurality of dimensions, the partial distance elimination occurring when less than all of the plurality of dimensions are sequentially added; wherein the evaluating comprises maintaining a bound for the plurality of log likelihoods, the bound being determined by comparing the plurality of log likelihoods of Gaussians to a threshold value.