Spatial voice activity detection using a microphone array noise canceller
Michael W. Hoffman, Zhao Li · The Journal of the Acoustical Society of America · 1999
An array of microphones is used for both noise suppression and spatial voice activity detection (SVAD). A generalized sidelobe canceller (GSC) is employed for adaptive interference rejection and the signals in the noise canceller are also exploited to provide desired signal activity detection. The SVAD shows promising results for environments with moderate reverberation and single or multiple interfering sources. For direct-to-reverberant energy ratios of 4 to 11 dB and signal-to-interference power ratios of −10 to 10 dB, speech detection miss rates are generally below 5%, while false alarm rates range from 10% to 40%. Most scenarios generally resulted in the false alarm rates between 10% and 20%. A key advantage of the GSC-based SVAD is its computational simplicity. In addition, the proposed technique is quite suitable for multiple interfering sources as well as nonstationary interference such as competing speakers. The system has been tested with a standard CELP speech coder.