基于小波支持向量机的传感器非线性校正

Nonlinear Emendation for Sensor Based on Wavelet Kernel Support Vector Machine

结合支持向量机和小波理论的优点,提出了一种基于小波核支持向量机的传感器非线性误差校正的原理和方法。该方法利用小波的多尺度插值特性和稀疏变化特性,提高了支持向量机的泛化能力和抗噪声能力。将该方法用于电涡流传感器的非线性校正,实验结果表明该小波核方法的校正效果优于传统的多项式拟合方法和RBF核支持向量机,提高了电涡流传感器测量的准确性。

Combining the advantages of support vector machine （SVM） and wavelet theory, the paper brings up a sort of sensor non- linear calibration method which bases on wavelet-kernel SVM. Making use of the characteristics of wavelet multi-scale interpolation and sparse variation, this method has improved the generalization ability and the noise resisting ability of vector support machine. The method used for eddy current sensor nonlinear correction, experimental results show that the correction effect of wavelet kernel meth- ods is superior to the traditional polynomial fitting methods and RBF kernel support vector machine, greatly increasing the eddy current sensor measurement accuracy.