On an Advanced Strategy of Gammachirp Wavelet Transform for Isolated Words Recognition using HMM

The gammachirp filter bank was used to model human cochlea filtering. Recently, we have proposed an auditory feature based on gammachirp wavelet transform. We have found that the features extractor gives a more significant improvement in robust speaker recognition than conventional acoustic features. The Gammachirp Wavelet Transform Frequency Cepestral Coefficients GWTFCC proposed the use of the Gammachirp Wavelet Transform in place of the Mel conventional MFCCs frame work. A comparative study shows that proposed features perform substantially better than conventional Mel-frequency cepstral. Keywords— Gammachirp, Wavelet Transform, Hidden Markov Models, Mel-Frequency Cepestral Coefficients.