Speech classification in noisy environment using subband decomposition

This paper presents a new algorithm for robust speech classification in adverse conditions, using an appropriate wavelet packet decomposition of the speech signal. The classification is achieved by generating a correlation functions of different subbands signals derived from a tree structured filter banks. The performance of the proposed technique is evaluated on speech signal (timit database) with real world noise, added to it, at various SNR. Experimental results show the accuracy of the proposed technique especially in low signal to noise circumstance (≤ 10 dB).