Calibration of a Six-Port Position Sensor via Support Vector Regression

In this paper, a calibration technique for the position sensor via support vector regression (SVR) is proposed. The position sensor adopts a zero-intermediate frequency architecture based on a six-port network, which is used for directly measuring the phase differences and indirectly reflecting the position. The SVR, which implements the structural risk minimization (SRM) principle, provides a good generalization ability from size-limited data sets. The results indicate that the SVR model can achieve a great predictive ability in positioning, with an accuracy of 2.41 mm over a distance range of 274.5 mm.