Improving speaker identification system using discrete wavelet transform and AWGN

This paper presents a robust speaker identification method from degraded noisy speech signals. This method is based on Mel-Frequency Cepstral Coefficients (MFCCs) for feature extraction from the noisy speech signals and Discrete Wavelet Transform (DWT). A comparative analysis is carried out with the traditional MFCCs based feature extraction method from noisy speech signals with additive white Gaussian noise (AWGN). The implementation mainly incorporates MFCCs which used for feature extraction and Vector Quantization using the Linde-Buzo-Gray (VQLBG) algorithm. It aims to minimize the amount of data to be handled. Results show that feature extraction from DWT of the degraded signals adds more speech features from the approximation and detail components. This helps achieving higher identification rates. Results also show that the proposed method improves the recognition rates computed at different degradation levels using different values of SNR cases.