Speech signal recognition based on genetic algorithm and Fisher projection

Based on the dynamic characteristic of speech signal, we proposed a new method of speech recognition to solve the optimal discriminant basis using genetic algorithms (GA) and Fisher projection. New optimal eigenvector separability was obtained by projecting the original eigenvector to the optimal discriminant basis. Six Chinese vowels were taken as the experimental data, and the MFCC coefficients, sub-band energy ratio of wavelet transform, pitch frequency, formant frequency and zero-crossing frequency of speech signal were taken as the original eigenvector, Then the suboptimal eigenvector was found out from the original one by GA, projecting the suboptimal eigenvector selected by GA to optimal discriminant basis. Finally using chaos neural network as the classifier, experiments show that the chaos neural network has preferable classification performance with optimal discriminant features.