BriefGPT.xyz
Oct, 2019
端到端语音分离的滤波器设计
Filterbank design for end-to-end speech separation
HTML
PDF
Manuel Pariente, Samuele Cornell, Antoine Deleforge, Emmanuel Vincent
TL;DR
本文提出复数解析滤波器及其相关表征与掩码策略，进一步拓展了单通道语音分离的滤波器发展，并展示其在嘈杂环境下的优秀性能。同时，基于本文提出的参数化滤波器，对所有情况均证明了复数表征与掩码的优势，并表明了短时傅里叶变换（STFT）在2ms窗口下具有最佳性能。
Abstract
Single-channel
speech separation
has recently made great progress thanks to learned
filterbanks
as used in ConvTasNet. In parallel, parameterized
→