Align-ULCNet: Towards Low-Complexity and Robust Acoustic Echo and Noise Reduction

2024-10-17

arXiv.org

AI Summary

To address the dual requirements of low computational complexity and high robustness for acoustic echo and noise reduction (AENR) on consumer-grade devices, this paper proposes a lightweight and efficient architecture. Methodologically, it introduces a novel time-aligned parallel encoder fusion structure coupled with a channel-wise feature redirection mechanism to enable robust cross-scenario modeling; further, it employs a hybrid temporal alignment strategy and a streamlined neural network design. The proposed approach significantly reduces computational and memory overhead while achieving superior echo suppression performance compared to existing state-of-the-art (SOTA) methods and attaining current-best noise suppression results. Contributions include a real-time, generalizable, and deployment-feasible AENR solution tailored for edge devices—establishing a new paradigm for on-device AENR.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

The successful deployment of deep learning-based acoustic echo and noise reduction (AENR) methods in consumer devices has spurred interest in developing low-complexity solutions, while emphasizing the need for robust performance in real-life applications. In this work, we propose a hybrid approach to enhance the state-of-the-art (SOTA) ULCNet model by integrating time alignment and parallel encoder blocks for the model inputs, resulting in better echo reduction and comparable noise reduction performance to existing SOTA methods. We also propose a channel-wise sampling-based feature reorientation method, ensuring robust performance across many challenging scenarios, while maintaining overall low computational and memory requirements.

Problem

• Develop low-complexity acoustic echo and noise reduction solutions
• Enhance ULCNet with time alignment and parallel encoder blocks
• Ensure robust performance with low computational and memory costs

Innovation

• Hybrid approach enhances ULCNet model
• Integrates time alignment and parallel encoder
• Channel-wise sampling ensures robust performance