An error separation module with variable momentum algorithm for narrowband active noise control

Abstract

Narrowband active noise control (NANC) is effective for stationary low-frequency noise but struggles with non-stationary noise from rotating or reciprocating machinery whose speed changes cause frequency and amplitude variations. This leads to inaccuracies in the reference frequency estimation and reduced NANC performance. This paper introduces an error separation module (ESM) that integrates autoregressive (AR) models with a variable momentum least mean square (VMLMS) algorithm. This combination effectively separates and suppresses the NANC system residual error across frequencies, enhancing the system's ability to track frequencies accurately in non-stationary conditions. The proposed solution has been validated through extensive simulations and real noise reduction experiments, showing improved noise reduction performance, faster convergence, and a decrease in steady-state mean squared error (MSE).