一种传感器特性的高精度拟合方法

Fitting Sensor Characteristics with High Accuracy

摘要:

为提高传感器非线性特性的拟合精度,提出了一种基于最小二乘支持向量机(LSSVM)与量子粒子群优化算法(QPSO)的传感器特性拟合方法；该方法采用最小二乘支持向量机构建传感器特性的非线性回归模型,模型的参数向量由量子粒子群优化算法和学习样本平均绝对误差最小的准则进行优化；实验结果验证了该方法的有效性,其拟合绝对误差在10-9~10-7%之间,其拟合性能明显优于常规方法。

Abstract:

To improve the fitting accuracy of nonlinear characteristics of the sensor characteristics, a new method is proposed based on the integration of least squre support vector machine(LSSVM) and quantum-behaved particle swarm optimization algorithm(QPSO). Firstly, LSSVM is used to build the regression model of sensor characteristics.Then parameter vector of fitting model is intelligently tuned by QPSO and the principle of mean maximum absolute error minimization. The experimental results show that the absolute error is between 10-9~10-7%, which means that the proposed method in this paper is effective, and performance of fitting model is superior to ordinary methods.