Title Of Paper:
 
 
Improving Text Independent Speaker Identification by Using Genetic Algorithm for Feature Selection
Author's Name :  Hasti Baharipour, Mohammad Mosleh, Mehdi KhalfehNilsaz
KeyWords:  Text independent speaker identification, Feature selection, Genetic algorithm, Gaussian Mixture Model.
Pages:  38 -46
Volume: 2
Issue: 10
Year: 2014
Abstract:
 

Nowadays, there exist many ways for speaker identification using different classifiers. But the problem is that all these methods use numerous and sometimes repetitive and unsuitable features that increase the modeling costs and decreases the accuracy of speaker identification. So, we seek a way which has the ability to select the best subset of the extracted features and to increase the accuracy of the classifier for speaker identification. In this paper, we have proposed a method based on synergy Genetic evolutionary algorithm with Gaussian Mixture Model classifier considering the selected suitable features of speech to improve the efficiency of text independent speaker identification systems. The algorithm was examined using a sample consisting of 40 people ranging from 30 to 50 years of age who had been selected randomly from Fars Data base. Experimental results showed that the proposed algorithm compared with the GMM classifier, increased the mean accuracy of speaker identification to 11.37% and decreased the mean number of selected features to 53.84%.

Full Text: