HiFi-GAN: High-Fidelity Denoising and Dereverberation Based on Speech Deep Features in Adversarial Networks

10 Jun 2020 Jiaqi Su Zeyu Jin Adam Finkelstein

Real-world audio recordings are often degraded by factors such as noise, reverberation, and equalization distortion. This paper introduces HiFi-GAN, a deep learning method to transform recorded speech to sound as though it had been recorded in a studio... (read more)

PDF Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Speech Enhancement DEMAND Noisy/unprocessed PESQ 1.97 # 17
Speech Enhancement DEMAND HiFi-GAN PESQ 2.94 # 8

Methods used in the Paper