A novel multi-gradient direction FxLMS algorithm with output constraint for active noise control

In active noise control (ANC) system, the maximum output power is restricted by the actuator drive performance. In the case that the output signal power exceeds this limitation, the noise reduction performance of the ANC system will be deteriorated due to the inherent nonlinearity of the actuators, and the adaptive algorithm also will be degraded. Therefore, a multi-gradient direction filtered-x least mean square (MGD-FxLMS) algorithm with output constraint is presented. The main idea is, when the output signal breaches the constraint, to slightly adjust the weight vector if the reference signal has an obvious effect on the output signal, otherwise to adjust the weight vector substantially. Therefore, we add two base points, which are determined by the output signal and the power of the reference signal, and three gradient directions to better adjust the weight vector. Furthermore, these improvements can reduce the weight fluctuation and precisely adjust the output signal within the constraint. In comparison with the conventional method, this strategy improves the noise reduction performance and stability, but the computational cost increases slightly. The enhanced performance is verified by the simulation results with four different situations.