SCA: Streaming Cross-attention Alignment for Echo Cancellation

TL;DR

This paper introduces SCA, an end-to-end streaming cross-attention network for echo cancellation that handles unaligned inputs and simplifies the pipeline, outperforming existing methods on standard datasets.

Contribution

The paper presents a novel end-to-end echo cancellation model with streaming cross-attention that does not require external alignment, improving robustness and simplicity.

Findings

Outperforms some hybrid and end-to-end methods on ICASSP2022 and Interspeech2021 datasets.

Handles unaligned inputs without external alignment.

Simplifies the echo cancellation pipeline for time-variant echo paths.

Abstract

End-to-End deep learning has shown promising results for speech enhancement tasks, such as noise suppression, dereverberation, and speech separation. However, most state-of-the-art methods for echo cancellation are either classical DSP-based or hybrid DSP-ML algorithms. Components such as the delay estimator and adaptive linear filter are based on traditional signal processing concepts, and deep learning algorithms typically only serve to replace the non-linear residual echo suppressor. This paper introduces an end-to-end echo cancellation network with a streaming cross-attention alignment (SCA). Our proposed method can handle unaligned inputs without requiring external alignment and generate high-quality speech without echoes. At the same time, the end-to-end algorithm simplifies the current echo cancellation pipeline for time-variant echo path cases. We test our proposed method on the ICASSP2022 and Interspeech2021 Microsoft deep echo cancellation challenge evaluation dataset, where our method outperforms some of the other hybrid and end-to-end methods.