BriefGPT.xyz
May, 2019
MetricGAN:基于生成对抗网络的黑盒度量分数优化应用于语音增强
MetricGAN: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement
HTML
PDF
Szu-Wei Fu, Chien-Feng Liao, Yu Tsao, Shou-De Lin
TL;DR
本文提出了一种基于MetricGAN的条件生成对抗网络，针对一种语音增强任务作了实验，该方法可优化生成器以提高被用户指定的一项或多项评估度量标准的得分。
Abstract
Adversarial loss in a
conditional generative adversarial network
(GAN) is not designed to directly optimize
evaluation metrics
of a target task, and thus, may not always guide the generator in a GAN to generate d
→