METHOD FOR UPGRADING SPEED OF VOICE RECOGNITION, DEVICE AND COMPUTER-READABLE MEDIUM

摘要

A method to improve a speed of voice recognition according to an embodiment of the present invention comprises: a step of extracting a voice characteristic vector from inputted voice data; a step of displaying the voice characteristic vector on a codebook and calculating distances between each of a plurality of cells and cells having the voice character vector displayed thereon; a step of selecting cells of which distances from the cells having the voice character vector displayed thereon are below a threshold; a step of calculating an observation probability of the voice character vector based on a Gaussian component of the selected cells; a step of calculating a language probability of the voice character vector by using calculated language model data; and a step of performing voice recognition by using the calculated observation probability and language probability.