一种基于BLMS的Volterra级数简化辨识方法

Simplified Volterra Series Identification Method Based on BLMS Filter

针对Volterra级数辨识中的“维数灾难”问题,提出了一种基于分块最小均方(BLMS)滤波器的简化辨识方法。该方法利用影响指数的概念,在保证一定辨识精度的前提下,根据每个Volterra核函数对辨识结果贡献的大小对其进行筛选,然后用筛选出的有效核函数作为对原系统的近似,从而达到降低辨识维数减少计算量的目的。该方法为克服“维数灾难”,实现在线辨识Volterra级数提供了一个有效解决途径。最后,文中用一个工程实例验证了该算法的有效性。

Volterra series identification often suffers from the so called ＇dimension disaster＇ because the scale of Volterra kernels expand exponentially with the increasing of system＇s order and degree. In order to solve this difficulty, a simplified identification method based on Block Least Mean Square （BLMS） filter was proposed. By means of this method, the concept of the effect index was introduced in order to reduce the scale of Volterra kernels by selecting them according to the degree of their effectiveness in identification. This method is suitable for Volterra identification on line and its effectiveness has been validated by an engineering instance.