Wavelet packet based voiced / unvoiced classification in noisy environment

This paper describes a new robust voiced/unvoiced classification algorithm, using an appropriate wavelet packet decomposition of the speech signal. The classification is achieved by generating a correlation model of different subbands signals derived from a tree structured filter banks. The wavelet packet tree is constructed by cascading the basic two channel perfect reconstruction filters into the desired levels. To investigate the accuracy of the proposed technique, we conduct experiments using the TI M IT speech database. We add to these speech signals real world noise at various SNR. Experimental results show the accuracy of the proposed technique especially in low SNR's (≤ 10dB).