多维次成分并行提取算法

Parallel Algorithm for Multiple Minor Component Extraction

针对目前多维次成分提取算法限制条件多和初始参数难以选择问题,在研究Douglas次子空间算法基础上,基于加权矩阵法提出了一种新型多维次成分并行提取算法。对该算法的自稳定性和收敛性分析表明:在输入信号有界和学习因子足够小时,该算法状态矩阵的模值总能收敛至一个常数;当且仅当状态矩阵收敛至需提取的多维次成分时,该算法达到稳定状态。仿真试验表明,与现有算法相比,该算法具有参数选取方法简单、易于实现和收敛速度快的优点。

The existing multiple minor component extraction algorithms have two problems, in cluding requiring too many limited conditions and choosing the initial parameters to be difficult. Aimed at the above problems, based on studying Douglas minor subspace tracking algorithm, a novel parallel algorithm for multiple minor component extraction is proposed by the weighted ma trix method. The analysis on its self-stability and the convergence shows that with the boundary input signal and a small enough learning factor, the state matrix module value of the algorithm always converges a constant. It also shows that if and only if the state matrix converges to the de sired multiple minor components, the algorithm can achieve the stability status. The simulation experiment shows that the algorithm has advantages of simpler parameter choose method, easier to realize and faster convergence speed than the existing algorithms.