A Decoupling Algorithm Based on PID Neural Network for Multi-Channel Active Noise Control of Nonstationary Noise

Abstract

It is critical to study multi-channel active noise control (ANC) systems to satisfy the requirements of noise reduction in multi-target positions. However, the complexity of the system may increase as the result of an increased number of sound channels. In addition, multi-channel coupling affects the stability of a system. In this paper, a decoupling algorithm based on the Proportion Integration Differentiation (PID) neural network and the filtered-x least-mean-square (FxLMS) for the multi-channel ANC of nonstationary noise is proposed. Due to the nonlinear characteristics of the PID neural network, the coupling problem can be solved through the algorithm. The performance of the novel proposed algorithm is verified by comparing the simulation results with the results from the traditional FxLMS algorithm and the FxLMS algorithm with matrix decoupling. The results illustrate that the convergence speed of the traditional FxLMS algorithm is similar to that of the FxLMS algorithm with matrix decoupling, while the proposed algorithm converges significantly faster than the other two algorithms. In terms of control performance, the proposed algorithm executes the best and the residual error signal has the minimum amplitude, followed by the FxLMS algorithm with matrix decoupling and the traditional FxLMS algorithm. With the advantages in convergence speed and control performance, the proposed decoupling algorithm could be suitable for multi-channel nonstationary ANC.