Wavelet-based multi-feature voiced/unvoiced speech classification algorithm

A new wavelet-based multi-feature voiced/unvoiced speech classification algorithm is presented. The algorithm is based on statistical analysis of wavelet-based frequency distribution of the average energy, zero-crossing rate, and average energy of short-time segments of the speech signal. The algorithm first classifies the input speech into voiced, unvoiced and uncertain parts by comparing features with predetermined thresholds. Then, the uncertain parts are treated in three conditions and dynamic thresholds are computed by extracted features of the input signal. Finally, the dynamic thresholds are used to classify the uncertain parts. The performance of the algorithm has been evaluated using a large speech database. The algorithm is shown to perform well in the cases of both clean and noise-degraded speech.