Adaptive Noise Cancellation Using Deep Cerebellar Model Articulation Controller

TL;DR

This paper introduces a deep cerebellar model articulation controller (DCMAC) for adaptive noise cancellation, leveraging a multi-layer structure to better model nonlinear transformations and improve noise reduction performance.

Contribution

The paper develops a novel DCMAC architecture with a modified backpropagation algorithm, extending traditional CMAC for enhanced nonlinear modeling in ANC applications.

Findings

DCMAC outperforms conventional CMAC in residual noise reduction

Deep structure enables better characterization of nonlinear transformations

Experimental results demonstrate improved ANC performance

Abstract

This paper proposes a deep cerebellar model articulation controller (DCMAC) for adaptive noise cancellation (ANC). We expand upon the conventional CMAC by stacking sin-gle-layer CMAC models into multiple layers to form a DCMAC model and derive a modified backpropagation training algorithm to learn the DCMAC parameters. Com-pared with conventional CMAC, the DCMAC can characterize nonlinear transformations more effectively because of its deep structure. Experimental results confirm that the pro-posed DCMAC model outperforms the CMAC in terms of residual noise in an ANC task, showing that DCMAC provides enhanced modeling capability based on channel characteristics.