FADI-AEC: Fast Score Based Diffusion Model Guided by Far-end Signal for Acoustic Echo Cancellation

The paper proposes two novel models: DI-AEC and FADI-AEC for Acoustic Echo Cancellation (AEC). These models pioneer a diffusion-based stochastic regeneration approach for AEC, with FADI-AEC designed to save computational demands, making it favorable for edge devices. The method outperforms some of the end-to-end methods and other diffusion-based echo cancellation methods when tested on the ICASSP2023 Microsoft deep echo cancellation challenge evaluation dataset.

Publication date: 8 Jan 2024
Project Page: https://arxiv.org/abs/2401.04283v1
Paper: https://arxiv.org/pdf/2401.04283