Frequency-Weighted Training Losses for Phoneme-Level DNN-based Speech Enhancement.

Nasser-Eddine Monir, Paul Magron, Romain Serizel

Browse the full MMSP paper archive.