Experimental study on applying adaptation techniques for deep neural network based voice activity detection

초록

In this paper, we employ several model-space adaptation techinique for deep neural network (DNN) based voice activity detection (VAD) to adapt the model to unseen background noise conditions. Adaptation results are evaluated in terms of th e objective measures such as frame accuracy, speech hit rate (HIT), and false alarm rate (FA). Experimental results suggest that the adaptation has been carried out mainly to leearn the noise-specific characteristics, rather than modeling the speech related features of the unseen adaptation utterances.