Feature extraction algorithm fusing GFCC and phase information
Yi Zhang, Lei Ni · 2017
For the traditional Mel cepstral coefficient feature extraction method is not completely in line with the characteristics of human audition, an auditory feature extraction algorithm based on Gamma-chirp filter cepstrum coefficient and phase information is proposed. Firstly, GFCC and phase information can be extracted through Gama-chirp filter banks, and then we comply discrete cosine transform and Kernel principal component analysis to fuse the obtained parameters for the better speaker feature Through this processing, the lower dimension of feature parameters can be received, noise robust performance of the recognition system can be improved as well. Experimental results show that the performance of the proposed method is much more better than others in the aspect of recognition rate and noisy-robust.