VB-DemandEx
홈페이지 · 논문 1편
Uses same clean speech as VoiceBank+Demand but more noise types. Features much lower SNRs ([−10, −5, 0, 5, 10, 15, 20] dB) compared to VoiceBank+Demand and has babble and speech shaped noise in both train, validation, and test set.
Speech English