An High Efficiency Feature Extraction Based on Wavelet Transform for Speaker Recognition

Abstract

To speaker recognition problem, firstly this paper will study and compare to various feature extraction methods include LPCC, PCA, fractal, and wavelet transform, which combined probabilistic neural network classifier. We carry out a set of experiments in speaker identification and matching .The result reveals Fractal has the best efficiency and discrete wavelet transform has the excellently high recognition rate. Besides, we will apply wavelet transform to reduce data dimension and enhance discriminative feature in speech signal, and combine LPCC, PCA, or fractal for feature extraction. The advantage of these mixed methods has the discriminative features in speaker recognition, saving system resource and speeding up recognition time. From our speech database, the average recognition of WT+LPCC in 10 times tests is 99.5% and the EER of speaker matching is 0.0. This shows the feature extraction method is combined with wavelet has excellently efficiency and performance.