WHAMR!: Noisy and Reverberant Single-Channel Speech Separation

TL;DR

This paper introduces WHAMR!, an extension of the WHAM! dataset that includes reverberation to better simulate real-world noisy indoor environments, and analyzes how current speech separation methods perform under these conditions.

Contribution

The paper presents WHAMR!, a new dataset with reverberated speech, and provides baseline evaluations of existing separation techniques on this more realistic data.

Findings

Reverberation significantly degrades separation performance.

Current methods struggle with reverberant conditions.

Cascaded architectures show potential improvements.

Abstract

While significant advances have been made with respect to the separation of overlapping speech signals, studies have been largely constrained to mixtures of clean, near anechoic speech, not representative of many real-world scenarios. Although the WHAM! dataset introduced noise to the ubiquitous wsj0-2mix dataset, it did not include reverberation, which is generally present in indoor recordings outside of recording studios. The spectral smearing caused by reverberation can result in significant performance degradation for standard deep learning-based speech separation systems, which rely on spectral structure and the sparsity of speech signals to tease apart sources. To address this, we introduce WHAMR!, an augmented version of WHAM! with synthetic reverberated sources, and provide a thorough baseline analysis of current techniques as well as novel cascaded architectures on the newly introduced conditions.