Self-Boosted Weight-Constrained FxLMS: A Robustness Distributed Active Noise Control Algorithm Without Internode Communication

Junwei Ji, Dongyuan Shi, Zhengding Luo, Boxiang Wang, Woon-Seng GanIEEE Signal Processing Letters 2025 · DOI

Introduction

Active Noise Control (ANC) is a critical technology for reducing acoustic pollution in environments such as office spaces and vehicle cabins. By generating "anti-noise" signals that are 180 degrees out of phase with primary noise sources, ANC systems can significantly attenuate unwanted sound. However, scaling these systems to large environments requires Multichannel ANC (MCANC) configurations, which face a significant technical trilemma.

Centralized systems, while stable and effective, suffer from high computational costs that scale poorly with the number of channels. Conversely, decentralized systems are computationally efficient but often fail due to the "crosstalk effect," where the anti-noise from one node is perceived as additional noise by another, leading to system divergence. Distributed systems attempt to bridge this gap through internode communication, but this introduces hardware complexity and latency issues that can compromise real-time performance. This paper presents the Self-Boosted Weight-Constrained Filtered-Reference Least Mean Square (SB-WCFxLMS) algorithm, a distributed solution that achieves centralized-level performance and high stability without requiring any internode communication.

Methodology

The proposed SB-WCFxLMS algorithm introduces two key mechanisms to ensure stability and performance in a distributed environment: Weight-Constrained FxLMS (WCFxLMS) and a Self-Boosted strategy.

Weight-Constrained FxLMS (WCFxLMS)

The core of the WCFxLMS approach is the addition of a penalty term to the standard FxLMS cost function. This term acts as a "tether," preventing the filter weights from deviating excessively from a reference center point ($\mathbf{\bar{w}}_k$). By constraining the weights, the algorithm effectively limits the node's sensitivity to acoustic crosstalk from neighboring nodes. This tethering ensures that even when nodes operate independently, they do not enter the feedback loops that typically lead to instability in decentralized systems.

Self-Boosted Strategy

To prevent the fixed constraint from limiting the system's ultimate noise reduction capability, the authors implemented a "self-boosted" mechanism. This module periodically monitors the local Residual Noise Level (RNL). If the current filter configuration demonstrates superior performance compared to previous states, the algorithm updates the constraint center point ($\mathbf{\bar{w}}_k$) to the current weight vector. This allows each node to independently "explore" and lock into more effective noise-canceling states, effectively boosting performance over time without central coordination.

Architecture of the SB-WCFxLMS node, featuring the WC-FxLMS controller and the Self-boost module for local optimization.

Experimental Evaluation

The researchers validated the SB-WCFxLMS algorithm using numerical simulations based on real-world acoustic data collected in a noise chamber with a 6-node ANC window setup. The system was tested against three primary noise types:

  1. Multi-tonal noise: Frequencies at 300, 400, 500, 600, and 700 Hz.
  2. Real-recorded compressor noise: Representing complex, non-stationary industrial sounds.
  3. Broadband noise: Focused on the 200–600 Hz range.

Key Results and Discussion

The experimental results demonstrate that SB-WCFxLMS successfully addresses the stability issues inherent in decentralized systems while matching the performance of centralized architectures.

Stability and Convergence

A primary finding is the superior stability of the proposed method. In scenarios where standard decentralized algorithms (DeMCANC) diverged and failed due to crosstalk, SB-WCFxLMS remained stable and achieved consistent noise reduction. As shown in the performance comparisons, the SB-WCFxLMS algorithm converges reliably, whereas competing decentralized methods often "blow up" shortly after activation.

Noise reduction performance (ANSE) over time, comparing SB-WCFxLMS against centralized and decentralized baselines. Note the divergence of DeMCANC.

Performance Benchmarking

In multi-tonal noise tests, the SB-WCFxLMS algorithm achieved noise reduction levels nearly identical to the computationally intensive centralized MCANC. Furthermore, when tested against recorded compressor noise, the method significantly outperformed the "Leaky FxLMS" algorithm—a common industry standard for stability—by providing deeper and more robust noise attenuation.

Conclusion

The SB-WCFxLMS algorithm represents a significant advancement in distributed active noise control. By utilizing local metrics and a weight-constrained update rule, the system eliminates the need for internode communication, thereby reducing hardware complexity and latency. The results confirm that local optimization, when properly constrained, can lead to global system stability and high-performance noise cancellation, making it a viable solution for large-scale, real-world ANC applications.

Summary updated

BibTeX
@article{ji2025self,
  author = {Junwei Ji and Dongyuan Shi and Zhengding Luo and Boxiang Wang and Woon-Seng Gan},
  title = {{Self-Boosted Weight-Constrained FxLMS: A Robustness Distributed Active Noise Control Algorithm Without Internode Communication}},
  journal = {IEEE Signal Processing Letters},
  year = {2025},
  doi = {10.1109/lsp.2025.3588421},
}
Figure