New Adaptive Momentum Algorithm Boosts Distributed Noise Control Speed
Researchers use cosine similarity to dynamically adjust momentum, slashing convergence time.
Distributed multichannel active noise control (DMCANC) spreads processing across multiple nodes, reducing the computational load of centralized systems but requiring information exchange that can suffer from communication delays. The existing auto-shrink step size mixed-gradients filtered reference LMS (ASSS-MGDFxLMS) algorithm improves robustness to delays but slows convergence due to reduced step sizes.
To overcome this, the team proposes the adaptive momentum ASSS-MGDFxLMS (AMAS-MGDFxLMS) algorithm. It introduces an adaptive momentum term that evaluates alignment between the instantaneous gradient and the momentum component using cosine similarity. When directions align, momentum is increased to accelerate convergence; when misaligned, it is reduced to maintain stability. Simulation results confirm the method achieves faster convergence than ASSS-MGDFxLMS while preserving effective noise reduction, even under delayed communication conditions. This work has practical implications for real-time noise cancelation in smart environments, industrial settings, and audio systems where network delays are unavoidable.
- Introduces adaptive momentum using cosine similarity to dynamically adjust momentum parameter.
- Achieves faster convergence than ASSS-MGDFxLMS under communication delays.
- Maintains stable and effective noise reduction performance in simulations.
Why It Matters
Faster, delay-robust noise cancellation enables real-time active noise control in distributed IoT and industrial audio systems.