基于数据驱动的非线性有源噪声MFFsLMS算法

MFFs LMS algorithm for nonlinear active noise control based on data driven

研究了非线性有源噪声控制(ANC)问题。采用函数连接型人工神经网络,以勒让德多项式作为扩展函数,提出了基于数据驱动的无模型滤波-s最小均方(MFFs LMS)算法。采用同步扰动随机逼近算法,估算系统的输出误差梯度。有效解决了因次路径时变所引起的系统稳定性问题。在理论分析的基础上,对该算法进行了仿真研究。仿真结果表明,当系统中呈现非线性及时变特性时,该方法能有效地抑制噪声且对系统次路径的变化具有良好的鲁棒性。

The nonlinear active noise control（ ANC） is studied. Using the functional link artificial neural network and Legendre polynomials extension function,the model freeed filter-s least mean square（ MFFs LMS） algorithm based on data driven is presented. The simultaneous perturbation stochastic approximation algorithm is used to estimate the gradient of the output error,which solves the stability problem caused by the time-varying characteristics of the secondary acoustic path. Based on theoretical analysis,computer simulations are carried out to validate the effectiveness of the proposed algorithm.Simulation results show that the proposed algorithm has good noise-canceling performance when the ANC system exhibits nonlinear,and is robust against the change of ANC secondary acoustic path.