参数自适应选择的局部投影滤波算法研究

Improvement of the Local Projective Algorithm for Signal Filtering Based on the Adaptive Parameter Selection

针对常规局部投影滤波算法中对邻域半径及噪声子空间维数的选取问题,提出了一种参数自适应选择的局部投影改进算法。利用相点与其近邻点形成的空间矢量方向随邻域半径的变化趋势,自适应地选取最优邻域半径;并采用MInka Beyasian模型选择(MInka Beyasian Selection,MI-BS)准则确定该邻域内噪声空间维数的大小。对Henon映射序列及实际采集的碰摩转子振动信号序列进行的仿真实验结果说明,该自适应滤波算法能够更精确地识别出噪声中的混沌数据,从而具有更强的混沌信号恢复及非线性降噪能力。

Aiming at the improper selection of neighborhood diameter and subspace dimension of noise in normal lo- cal projective algorithm, an improving method for adaptive parameter selection is proposed. Using the curve dia- gram of the space vector direction which formed by phase point and its neighbor points within the whole given neigh- borhood according to the varying diameters, the optimum neighborhood diameter will be appropriately fixed. And the dimension of noise subspace is exactly determined by MInka Beyasian Selection （MIBS） criterion. The result of simulation experiments on Henon sequence and actual vibration signals of rub-impact rotor indicates that the adap- tive filtering algorithm has the better capability of rebuilding chaotic signals and reducing nonlinear noises in the way of more precise reorganization between chaotic data and noise.