Coevolutionary approach to speaker identification using neural networks
Xuming He, Guangrui Hu, Tan Zheng-Hua · 2002
This paper introduces a coevolutionary neural network approach for design of speaker identification system. In the proposed training algorithm, neural networks for speaker modeling and feature subsets used for identification will be automatically designed by one evolution process in which a new hybrid evolutionary algorithm is employed. The experimental results show that new method achieves approximately 6% increase in recognition correct rate and uses less number of features in the classification.