一种空间分布式变阶数自适应网络滤波算法

现有的自适应网络滤波算法大都假设未知参数向量的阶数已知且恒定，无法解决阶数未知或时变条件下的参数向量估计问题. 以最小化网络均方误差为准则，提出一种空间分布式变阶数自适应网络滤波算法. 该算法仅要求网络中的各节点与相邻节点进行通信，通过扩散的方式实现整个网络数据信息的融合，具有计算量小、可操作性强及估计精度高的特点.仿真表明，提出的算法能够有效地估计和跟踪未知参数向量的阶数和权值.

A Spatially Distributed Variable Tap-length Strategy over Adaptive Networks

Among the existing strategies over adaptive networks, the tap-length of an unknown parameter vector is assumed to be known a prior and constant, thus they are not suitable in the context where the optimal tap-length is unknown or variable. We therefore propose a spatially distributed variable tap-length algorithm over adaptive networks, based on the rule of least mean square error of the entire network. In the approach, the data of the network diffuse across the nodes through local iterations between the nodes and their neighbors, thus the accuracy is guaranteed, while only a small computation is required. Simulations show that the proposed strategy is effective to track and estimate the parameter vector of interest in terms of tap-length and weights.