基于变中心互相关熵的比例自适应滤波算法研究

针对传统自适应滤波算法对于非零均值非高斯噪声干扰环境下稀疏系统参数估计存在稳态精度低的问题,以变中心互相关熵为代价函数,引入比例更新机制,应用随机梯度法设计一种新的稀疏自适应滤波算法。变中心互相关熵的中心可位于任何位置,其可很好地匹配非零均值的误差分布,而比例更新机制为每个权值参数赋予可变的步长参数,因此可增强算法的跟踪能力。进一步设计在线学习方法来估计核宽度和中心位置,以提高算法性能。另外根据能量守恒关系研究了算法的收敛性。仿真实验结果表明,该算法相对于传统自适应滤波算法对于非零均值非高斯噪声环境下的稀疏参数估计具有明显的优越性和鲁棒性。

Research on proportionate adaptive filter based on correntropy with variable center algorithm

In view of the low steady-state accuracy of the traditional adaptive filtering algorithm for the estimation of sparse system parameters under non-zero mean non-Gaussian noise interference environment,this paper designed a novel sparse adaptive filtering algorithm by using the correntropy with variable-center as the cost function and employing the proportional update mechanism.The center of the correntropy with variable center could be located at any position,which could well match the error distribution of non-zero mean,and the proportional update mechanism gave variable step parameters for each weight parameter,and thus enhancing the tracking ability of the algorithm.This method further used an online learning method to estimate the kernel width and center adaptively to improve the performance of the algorithm.In addition,this paper studied the convergence of the algorithm according to the energy conservation relation.The simulation results show that the proposed algorithm is superior and robust to the estimation of sparse parameters in non-zero mean non-Gaussian noise environment compared with traditional adaptive filtering algorithms.