Improved Speaker Identification Algorithm based on Discriminative Weighted Method

摘要

How to decrease false acceptance rate is an ever-lasting target in speaker identification system. This paper proposes a new method based on the thought of discriminative weighting, which is realized by previously distributing a discriminative weighted value vector between each two speakers. The improved algorithm means to calculate discriminative weighted values based on the difference between each two speakers, which are formed into a discriminative weighted vector to discriminate the similarity of test speech from these two speakers. In the identification process, test speech will be measured between each two speakers one by one, and the final result will be derived from syncretizing these results. Effect of the improved algorithm was proved by experimental results.