Online neural fusion of distortionless differential beamformers for robust speech enhancement

2025-10-28

AI Summary

Fixed beamformers exhibit poor adaptability in dynamic acoustic environments, while conventional adaptive convex combination (ACC) algorithms struggle to track rapidly time-varying interferences. To address these limitations, this paper proposes a frame-level online neural fusion framework. It employs a lightweight neural network to predict, in real time, the combination weights for multiple distortionless differential beamformers, enabling dynamic weighted fusion under strict distortionless constraints. This work constitutes the first integration of neural networks into online fusion of multiple beamformers, overcoming ACC’s tracking bottleneck in highly non-stationary scenarios—such as those involving fast-moving interferers. Experimental results demonstrate that the proposed method significantly outperforms traditional ACC under rapid interference variation, achieving a superior trade-off between speech fidelity and interference suppression. Consequently, it enhances both robustness and overall speech quality.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Fixed beamforming is widely used in practice since it does not depend on the estimation of noise statistics and provides relatively stable performance. However, a single beamformer cannot adapt to varying acoustic conditions, which limits its interference suppression capability. To address this, adaptive convex combination (ACC) algorithms have been introduced, where the outputs of multiple fixed beamformers are linearly combined to improve robustness. Nevertheless, ACC often fails in highly non-stationary scenarios, such as rapidly moving interference, since its adaptive updates cannot reliably track rapid changes. To overcome this limitation, we propose a frame-online neural fusion framework for multiple distortionless differential beamformers, which estimates the combination weights through a neural network. Compared with conventional ACC, the proposed method adapts more effectively to dynamic acoustic environments, achieving stronger interference suppression while maintaining the distortionless constraint.

Problem

• Enhancing robustness of beamforming in varying acoustic conditions
• Overcoming limitations of adaptive algorithms in non-stationary scenarios
• Improving interference suppression while maintaining distortionless speech quality

Innovation

• Neural network estimates beamformer combination weights
• Fusion framework adapts to dynamic acoustic environments
• Multiple distortionless differential beamformers enhance interference suppression