基于混合熵的分数阶归一化子带自适应滤波算法

Fractional Normalized Subband Adaptive Filtering Algorithm Based on Mixture Correntropy

为了改善非高斯噪声环境下归一化子带自适应滤波算法的滤波性能,引入了最大混合相关熵准则和分数阶微分的概念。一方面,利用最大混合相关熵准则的鲁棒性,有效地减小了异常噪声对算法性能的影响;另一方面,在权重更新的过程中,通过引入分数阶微分部分,并以加权的形式考虑数据整体信息,更准确地描述了实际系统,从而提高了算法的滤波性能,可将这一改进后的算法应用于非高斯冲击噪声和有色噪声环境下的系统辨识和非线性信道均衡。通过仿真实验结果可以看出,与已有的鲁棒算法相比,所提算法具有更强的鲁棒性和更高的系统跟踪和估计能力。

In order to improve the filtering performance of the normalized subband adaptive filter(NSAF)in a non-Gaussian noise environment,the maximum mixture correntropy criterion and fractional-order differentiation are applied to the NSAF algorithm.On the one hand,the robustness of the maximum mixture correntropy criterion is used to effectively suppress the effect of anomalous noise values on the performance of the algorithm.On the other hand,to describe the actual system more accurately,a fractional-order differentiation component is added to the weight update.The proposed algorithm is applied to system identification and nonlinear channel equalization in a non-Gaussian interference noise and colored noise.Simulation results show that the proposed algorithm has stronger robustness and better system tracking and estimation capability compared with existing robust algorithms.