Audio samples for "Enhancing Low-quality Voice Recordings using Disentangled Channel Factor and Neural Waveform Model"


Authors: Haoyu Li, Yang Ai, Junichi Yamagishi

Submitted to IEEE SLT 2021


Recording condition: ipad_livingroom

Raw audio WPE WPE+L Wavenet ED ED+CM FULL Linear-ISTFT Studio audio

Recording condition: ipadflat_office

Raw audio WPE WPE+L Wavenet ED ED+CM FULL Linear-ISTFT Studio audio

Recording condition: iphone_bedroom

Raw audio WPE WPE+L Wavenet ED ED+CM FULL Linear-ISTFT Studio audio



Acknowledgement:
We used DAPS dataset in our experiments
Universal WaveRNN was pretrained by Mr. Eren Gölge: https://github.com/mozilla/TTS/issues/221