Voice Activity Detection using Group Delay Processing on Buffered Short-term Energy

In this paper, we present an algorithm for Voice Activity Detection (VAD) in speech signals using the minimum phase group delay function. The proposed method considers a buffer consisting of contiguous frames of the given signal and computes the short-term energy (STE) for that buffer. By appending a surrogate signal to STE and viewing the resultant signal as a positive part of the magnitude spectrum of an arbitrary signal, the minimum phase group delay function is computed. The group delay is then noise compensated and median filtered. The regions having positive group delay values are classified as speech and those with negative values are classified as noise. Experimental comparisons with the G.729 Annexe B VAD algorithm demonstrates significantly better performance for the proposed method, revealing that the algorithm is robust to noise.