Voice activity detection using frame-wise model re-estimation method based on Gaussian pruning with weight normalization

Masakiyo Fujimoto, Shinji Watanabe, Tomohiro Nakatani