Genetic Algorithm-Based Adaptive Active Noise Control Without Secondary Path Identification

The conventional filter-x least mean square (FxLMS) is a prevalent control algorithm for active noise control (ANC). Nevertheless, its derivation relies on gradients, rendering the performance of ANC prone to the influence of local minimum problem. Besides, identifying the secondary path in advance is a requirement for implementing the conventional FxLMS algorithm. In this article, a novel adaptive ANC system using an online genetic algorithm (GA) is proposed by recasting ANC as an optimization problem. It reformulates the optimization criterion and modifies the GA to match the ANC system, so as to realize the online adaptive attenuation of noise. Additionally, employing real number encoding to combine filter coefficients into chromosomes not only avoids the laborious decoding operation associated with binary encoding, but also facilitates getting rid of the local minima problem. The proposed approach eliminates the need for pre-estimating the transfer function of the secondary path, which is necessary in the FxLMS algorithm. Moreover, an effective reinitialization strategy is designed to further handle unforeseen alterations in secondary path. As per the simulation results, the proposed approach is able to successfully mitigate the noise without the requirement of secondary path identification.