基于最小二乘支持向量机的传感器动态系统辨识方法

A Dynamic System Identification Method for Sensors Based on LS-SVM

在介绍和比较标准支持向量机(SVM)和最小二乘支持向量机(LS-SVM)原理的基础上,提出了一种利用LS-SVM模型进行传感器动态系统辨识的方法,并给出了相应的过程和算法。与标准SVM模型比较,该方法优点是明显的:(1)用等式约束代替标准SVM算法中的不等式约束;(2)将求解二次规划问题转化为直接求解线性矩阵方程,使得在相同条件下,系统辨识速度提高1～2个数量级,辨识误差降低50%。因此,LS-SVM模型速度快,抗噪声干扰能力强,更适合传感器动态系统建模。

Based on introducing and comparing standard support vector machine （SVM） and least squares support vector machine （ LS-SVM）, an identification method of sensors dynamic systems using LS-SVM model was given. The design steps and learning algorithm were also addressed. Compared with standard SVM model, there were some advantages of this method ： 1 ） the constraints of inequalities were replaced by equality-type constraints in LS-SVM. 2）the LS-SVM solution followed directly from solving a set of linear equations instead of quadratic pro- gramming. In the same condition, the speed of identification was 10 ～10^2 times than that of standard SVM method, while the error of identification was about 50% of the SVM method. As a result, the presented method is faster in learning speed, higher in accuracy, more robust in noise resistance. So it is more suitable for sensors dynamic system identification.