An improved statistical model-based VAD algorithm with an adaptive threshold

A voice activity detection (VAD) algorithm with a fixed threshold cannot follow fluctuations of signal and noise power level in a time-varying environment and the inability to adapt to a time-varying environment severely limits the VAD performance. Therefore, we need to employ an adaptive threshold with which the VAD will have enhanced performance even in a time-varying SNR environment. In this paper, we propose an improved statistical model-based VAD algorithm employing preprocessing and an adaptive threshold. We also perform extensive computer simulations to demonstrate performance improvement of the proposed algorithm under various noise environments, when compared to other algorithms.