Wavelet-Based Mel-Frequency Cepstral Coefficients for Speaker\n Identification using Hidden Markov Models

Mahmoud I. Abdalla, Hanaa S. Ali

arXiv (Cornell University) · 2010 · 21 citations · 0 references

DOIFull text

Open access

Concepts

Abstract

To improve the performance of speaker identification systems, an effective\nand robust method is proposed to extract speech features, capable of operating\nin noisy environment. Based on the time-frequency multi-resolution property of\nwavelet transform, the input speech signal is decomposed into various frequency\nchannels. For capturing the characteristic of the signal, the Mel-Frequency\nCepstral Coefficients (MFCCs) of the wavelet channels are calculated. Hidden\nMarkov Models (HMMs) were used for the recognition stage as they give better\nrecognition for the speaker's features than Dynamic Time Warping (DTW).\nComparison of the proposed approach with the MFCCs conventional feature\nextraction method shows that the proposed method not only effectively reduces\nthe influence of noise, but also improves recognition. A recognition rate of\n99.3% was obtained using the proposed feature extraction technique compared to\n98.7% using the MFCCs. When the test patterns were corrupted by additive white\nGaussian noise with 20 dB S/N ratio, the recognition rate was 97.3% using the\nproposed method compared to 93.3% using the MFCCs.\n