Voice recognition based on the theory of transmission wave and LSSVM

Abstract

Based on introduced the principle of the one-dimensional voice transmission parameters and least squares support vector machine (LSSVM), one-dimensional transmission parameters are classified by LSSVM. γ and σ are important kernel parameters for LSSVM, and the kernel parameters of LSSVM are selected by 5-fold cross-validation method to obtainbetter recognition results. Finally, by comparing the identification results based on LSSVM with that based on BP neural network algorithm, it is found that the method of this paper is better than BP algorithm. © 2011 Springer-Verlag Berlin Heidelberg.