Discriminating between voice and silence based on an adaptation of hidden Markov models, unsupervised.

摘要

PRESENTED AN ADAPTATION HMM, LEVEL OF JUDGMENT, discriminatory, unmonitored, based classification CONVERSATION-SILENCE. DETERMINED THE REGIONS OF SILENCE AND CONVERSATION USING A POINTER END OF CONVERSATION OR OBTAINED FROM THE DEVICE SEGMENTATION RECONOCEDOR in a first step. Then the discriminatory PROCEDURE USED BY LEARNING A GPD OR ANY OTHER DISCRIMINATORY learning algorithm, in conjunction with HMM-BASED DEVICE RECONOCEDOR to increase the discrimination between silence and conversation.