Deep Filter Estimation from Inter-Frame Correlations for Monaural Speech Dereverberation

TL;DR

This paper introduces IF-CorrNet, a novel deep learning architecture that leverages inter-frame correlations to estimate filters for speech dereverberation, improving robustness and generalization in real-world scenarios.

Contribution

The paper presents a correlation-based deep filter estimation method that outperforms traditional direct mapping approaches in reverberant speech dereverberation tasks.

Findings

Significant SRMR improvement on RealData in REVERB Challenge

Enhanced robustness against acoustic variability

Effective suppression of reverberation and noise

Abstract

Speech dereverberation in distant-microphone scenarios remains challenging due to the high correlation between reverberation and target signals, often leading to poor generalization in real-world environments. We propose IF-CorrNet, a correlation-to-filter architecture designed for robustness against acoustic variability. Unlike conventional black-box mapping methods that directly estimate complex spectra, IF-CorrNet explicitly exploits inter-frame STFT correlations to estimate multi-frame deep filters for each time-frequency bin. By shifting the learning objective from direct mapping to filter estimation, the network effectively constrains the solution space, which simplifies the training process and mitigates overfitting to synthetic data. Experimental results on the REVERB Challenge dataset demonstrate that IF-CorrNet achieves a substantial gain in the SRMR metric on RealData, confirming its robustness in suppressing reverberation and noise in practical, non-synthetic environments.