Y$^2$-Net FCRN for Acoustic Echo and Noise Suppression

Abstract

In recent years, deep neural networks (DNNs) were studied as an alternative to traditional acoustic echo cancellation (AEC) algorithms. The proposed models achieved remarkable performance for the separate tasks of AEC and residual echo suppression (RES). A promising network topology is a fully convolutional recurrent network (FCRN) structure, which has already proven its performance on both noise suppression and AEC tasks, individually. However, the combination of AEC, postfiltering, and noise suppression to a single network typically leads to a noticeable decline in the quality of the near-end speech component due to the lack of a separate loss for echo estimation. In this paper, we propose a two-stage model (Y 2 -Net) which consists of two FCRNs, each with two inputs and one output (Y-Net). The first stage (AEC) yields an echo estimate, which-as a novelty for a DNN AEC model-is further used by the second stage to perform RES and noise suppression. While the subjective listening test of the Interspeech 2021 AEC Challenge mostly yielded results close to the baseline, the proposed method scored an average improvement of 0.46 points over the baseline on the blind testset in double-talk on the instrumental metric DECMOS, provided by the challenge organizers.

Deep Residual Echo Suppression and Noise Reduction: A Multi-Input FCRN Approach in a Hybrid Speech Enhancement System

“…The second major question to investigate for this application is whether it is more beneficial to perform a direct estimation of the complex output target with the network (regression) or if the nice improvements of the magnitude-bounded complex mask-based procedure in [11,3] can be transferred to this task as well.…”

“…In this typical two-stage arrangement for speech enhancement, the application of a DNN as second stage (RES and NR) gained increasing attention with early investigations of feed-forward networks [1,2], convolutional networks bringing further improvements more recently [3,4], some even being fully synergistic with the first stage [5], and many more. In the meantime, also fully learned deep AEC approaches were proposed, where a single network incorporates the tasks of AEC, RES, and NR, e.g., [6,7] or further investigated in [8].…”

“…Our contributions are as follows: First, we propose a multi-input FCRN for RES/NR and investigate all possible input signal combinations, revealing trade-offs between different performance aspects and providing surprising insights to simple design choices. Second, we investigate whether the improvements of the magnitude-bounded complex mask-based procedure in [11,3] can be transferred to this task as well. Third, with the right choices, we improve performance over state of the art.…”