An improved speech feature extraction algorithm using DWT

This paper proposes a novel approach for speech feature extraction., the proposed noise robust feature extraction algorithm is found to be robust to different noise characteristics, SNRs, and under various training-testing matched or unmatched conditions. And the algorithm this paper presents exhibit a low computational cost. We also give a practical weighting technique here. The experiment compares traditional MFCCs algorithm and our methods, results show overall improvement with the application of proposed wavelet based denoising algorithm with a modified thresholding procedure.