Spoken Word Recognition Using Statistical and Syntactic Methods in Multi talker Environment.

摘要

This work focuses on the recognition of spoken words in multi-talker environment. A new method of recognizing spoken words belonging to a given vocabulary spoken by a number of speakers, the identity of speaker being unknown to the system, is described and the experimental results reported. In the present work, the vocabulary has 40 words spoken by six speakers. The vocabularly consists of many pairs of words which are phonemically similar to each other. A recognition rate of 99 percent was obtained by testing about 400 word utterances. An attempt is made to use the same technique for recognizing phrases made up of 2-3 words.