Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

E2E-AEC: Implementing an end-to-end neural network learning approach for acoustic echo cancellation

Yiheng Jiang, Biao Tian, Haoxu Wang, Shengkui Zhao, Bin Ma, Daren Chen, Xiangang Li

TL;DR

This work tackles robust acoustic echo cancellation in streaming settings by proposing E2E-AEC, an end-to-end neural network that bypasses time delay estimation and traditional LAEC. The model leverages progressive learning, knowledge transfer from a pre-trained LAEC-based hybrid, an attention-based time alignment supervised by GCC-PHAT–derived delays, and VAD-driven masking to suppress echo during near-end silence. Key contributions include a unidirectional GRU–based architecture with a TF-GridNet–inspired design, explicit delay supervision for alignment, and layer-specific VAD masking that yields substantial improvements in ERLE on public AEC Challenge datasets. The results demonstrate strong MOS and echo suppression benefits, illustrating the practical potential of end-to-end approaches for real-world echo cancellation across diverse acoustic conditions.

Abstract

We propose a novel neural network-based end-to-end acoustic echo cancellation (E2E-AEC) method capable of streaming inference, which operates effectively without reliance on traditional linear AEC (LAEC) techniques and time delay estimation. Our approach includes several key strategies: First, we introduce and refine progressive learning to gradually enhance echo suppression. Second, our model employs knowledge transfer by initializing with a pre-trained LAECbased model, harnessing the insights gained from LAEC training. Third, we optimize the attention mechanism with a loss function applied on attention weights to achieve precise time alignment between the reference and microphone signals. Lastly, we incorporate voice activity detection to enhance speech quality and improve echo removal by masking the network output when near-end speech is absent. The effectiveness of our approach is validated through experiments conducted on public datasets.

E2E-AEC: Implementing an end-to-end neural network learning approach for acoustic echo cancellation

TL;DR

This work tackles robust acoustic echo cancellation in streaming settings by proposing E2E-AEC, an end-to-end neural network that bypasses time delay estimation and traditional LAEC. The model leverages progressive learning, knowledge transfer from a pre-trained LAEC-based hybrid, an attention-based time alignment supervised by GCC-PHAT–derived delays, and VAD-driven masking to suppress echo during near-end silence. Key contributions include a unidirectional GRU–based architecture with a TF-GridNet–inspired design, explicit delay supervision for alignment, and layer-specific VAD masking that yields substantial improvements in ERLE on public AEC Challenge datasets. The results demonstrate strong MOS and echo suppression benefits, illustrating the practical potential of end-to-end approaches for real-world echo cancellation across diverse acoustic conditions.

Abstract

We propose a novel neural network-based end-to-end acoustic echo cancellation (E2E-AEC) method capable of streaming inference, which operates effectively without reliance on traditional linear AEC (LAEC) techniques and time delay estimation. Our approach includes several key strategies: First, we introduce and refine progressive learning to gradually enhance echo suppression. Second, our model employs knowledge transfer by initializing with a pre-trained LAECbased model, harnessing the insights gained from LAEC training. Third, we optimize the attention mechanism with a loss function applied on attention weights to achieve precise time alignment between the reference and microphone signals. Lastly, we incorporate voice activity detection to enhance speech quality and improve echo removal by masking the network output when near-end speech is absent. The effectiveness of our approach is validated through experiments conducted on public datasets.
Paper Structure (14 sections, 5 equations, 2 figures, 3 tables)

This paper contains 14 sections, 5 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. METHODS
  3. Problem Formulation
  4. Overall Network Architecture
  5. Progressive Enhancement
  6. Knowledge Transfer
  7. Time Alignment
  8. VAD Prediction and Masking
  9. Loss Function
  10. EXPERIMENTS
  11. Data Preperation
  12. Evaluation Results
  13. CONCLUSION
  14. REFERENCES

Figures (2)

  • Figure 1: Figure 1. E2E-AEC system overview. The outputs include: delay (time delay estimation), vad (near-end speech VAD), and spec (spectrum estimations of different PL stages).
  • Figure 2: Figure 2. TDE results for a sample (ground truth delay: 650 ms ). Mean and variance are computed based on the difference between predictions and ground truth.