Entropy based Voice Activity Detection for VoIP

Voice Activity Detection (VAD) is used to detect thepresence of speech in an audio signal. VAD plays an importantrole as a preprocessing stage in numerous audio processingapplications. In Voice over IP (VoIP) and mobile telephonyapplications, VAD can reduce bandwidth usage and networktraffic by transmitting audio packets only if speech is detected.VAD for VoIP can save bandwidth by filtering the frame thatdoes not certain speech. VAD algorithm is used to identify thevoice and silent region of a speech. This paper uses aninformation theoretic measure, called spectral entropy fordifferentiating silence from speech zone. The performance of thesystem has been tested on signals with varying levels of SNRsin different environments. The result reveals that this system canfairly detect speech and non-speech events in noisybackgrounds