Voice Activity Detection Based on Sub-band Reprocessed Pectrum Entropy

Voice activity detection (VAD) in strong noise environments is improved by an algorithm based on subband reprocessed spectrum entropy (BRSE). As a new feature parameter for VAD, the reprocessed spectrum is calculated firstly, and reprocessed spectrum entropy is then calculated with multi-subband analysis. The order statistics filter is selected to smooth the BRPE. A new voice / noise discrimination algorithm is proposed by combining the finite state machine (FSM) with BRSE. The misdetections caused by single-threshold are reduced greatly. Experimental results show that the proposed algorithm has higher accuracy and stronger robustness than other two methods.