A Deep Recurrent Neural Network Controller For Nonlinear Active Noise Control Systems

The traditional adaptive filter and Filtered-x LMS (FxLMS) algorithm have been widely used in active noise control(ANC) systems. But for the nonlinearity in the system, the linear control algorithm will be difficult to get good control effect, while the neural network can overcome the influences of nonlinearity. In this paper, the deep recurrent neural network(DRNN) controller is applied to ANC systems to improve noise cancellation capability. Simulation results demonstrate that the proposed DRNN controller can obtain better performance than the FxLMS and Filtered-x BP Neural Network (FxBPNN) in the presence of strong nonlinearity.