New Filter Structure based on Admissible Wavelet Packet Transform for Text-Independent Speaker Identification

Identical acoustic features like Mel frequency cepstral Coefficients (MFCC)and Linear predictive cepstral coefficients (LPCC) are being widely used for different tasks like speech recognition and speaker recognition, whereas the requirement of speaker recognition is different than that of speech recognition. In MFCC feature representation, the Mel frequency scale is used to get a high resolution in low frequency region, and a low resolution in high frequency region. This kind of processing is good for obtaining stable phonetic information, but not suitable for speaker features that are located in high frequency regions. Further MFCC uses short time Fourier transform (STFT), which has fixed time-frequency resolution. Considering above facts, in this paper we have proposed a new filter structure based on admissible wavelet packet transform for text-independent speaker identification. Multiresolution capabilities of wavelet packet transform are used to derive the new features. The performance of the proposed features is evaluated using the most commonly used Gaussian mixture model (GMM) as well as the continuous density hidden Markov model (CDHMM) classifiers. Improved speaker identification rate is obtained using the proposed features compared to the MFCC and other Wavelet transform based features. Further the results show that CDHMM works better than the GMM for small number of mixture densities. Identification accuracy of 99.76% is achieved by conducting the experiments on TIMIT database.