Empirical Study of Decentralized Multi-Channel Active Noise Control Based on the Genetic Algorithm
* Presenting author
In an active noise control (ANC) system, computational complexity is one major concern when designing practical control algorithms. One approach to reducing computational complexity is to apply a decentralized control scheme rather than the centralized scheme. A decentralized scheme attempts to control a number of ANC subsystems independently, where for simplicity, one subsystem consists of one loudspeaker and one error microphone. Our recent published article has shown theoretically that decentralized two-channel ANC can achieve the same noise reduction performance as the centralized controller with guaranteed convergence in the frequency domain. In this work, we attempt to extend the results from two-channel case to N (N>1) channel case. The challenge sits in finding N complex numbers that could properly shape the eigenvalues of an N * N matrix for each frequency bin towards guaranteed convergence. Due to the problem complexity, we conduct empirical study by using the genetic algorithm (GA). Simulated results on the channel numbers of 2, 4, 6, and 12 demonstrate that the resulting decentralized ANC controller is also able to achieve the same noise reduction performance as the centralized controller.