Active noise control in free space using recurrent neural networks with EKF algorithm

This paper presents theoretical and experimental investigation of active noise control (ANC) in free space using recurrent neural networks. A learning algorithm for diagonal recurrent neural networks based on extended Kalman filter is developed and is referred to as diagonal recurrent extended Kalman filter (DREKF) algorithm. Based on DREKF, new control algorithm suited for ANC is developed to handle nonlinearity inherently arising in this application. Real-time experiment using floating point digital signal processor is carried out for both identification and control tasks required in ANC. The results show that the number of neurons in neural network can be reduced by introducing the diagonal recurrent elements, without deteriorating the system performance, and that DREKF produces better performance than linear adaptive controller in compensating the secondary path nonlinearity.