Wavelet Feature Selection using Genetic Algorithms for Text Independent Speaker Recognition

Shung-Yung Lung


Speaker recognition, Kernel canonical correlation analysis, Wavelet transform, Genetic algorithm


Kernel canonical correlation analysis as a powerful nonlinear feature extraction method has proven as a preprocessing step for classification algorithm. In this paper, a text independent speaker recognition system based on kernel canonical correlation analysis and genetic algorithms is proposed. By the use of the polynomial functions as kernel function in wavelet analysis, the high order relationships can be utilized and the nonlinear components can be obtained. After we got nonlinear components, we use genetic algorithms to select the optimal feature set for speaker identification. The experimental results on the TALUNG database and KING database illustrate the effectiveness of the proposed method.

Important Links:

Go Back