ViVoVAD: a Voice Activity Detection Tool based on Recurrent Neural Networks

Voice Activity Detection (VAD) aims to distinguishcorrectly those audio segments containing humanspeech. In this paper we present our latest approachto the VAD task that relies on the modellingcapabilities of Bidirectional Long Short TermMemory (BLSTM) layers to classify every frame inan audio signal as speech or non-speech