Mel-scaled Discrete Wavelet Transform and dynamic features for the Persian phoneme recognition

In this paper we use a feature vector consisting of the Mel Frequency Discrete Wavelet Coefficients to recognize spoken phonemes in the Persian language. The purpose of using wavelet in feature extraction is to benefit from its multi resolution analysis and localization property in time and frequency domains. The MFDWCs are obtained by applying the Discrete Wavelet Transform (DWT) to the Mel-scaled log filter bank energies of a speech frame. Feature vectors are used for the HMM-based phoneme recognition on a portion of the FarsDat Persian language database consisting of 35 hour recorded data for training and 15 hour for testing. We evaluate the performance of new features for clean speech and noisy speech and compare it with the Mel Frequency Cepstral Coefficients (MFCC). Experiments on a phone recognition task based on the MFDWC give better result than recognizers based on the MFCC features for both white noise and clean speech cases.