SVM multiple non-linear regression for moisture content detecting

A method for regression of non-linear relations between resonance parameters and moisture content is employed in order to eliminate the measurement errors. A multiple non-linear regression model based on support vector machine(SVM) is built. Then, the eigenvalue and contribution degree of resonance frequency, quality factor and environment temperature are calculated. Experiments are employed by SVM-KM toolbox with 50 group data for training model and 15 group data for verifying model performance. The result showed the arithmetic not only has the ability to realize the moisture soft-sensor using microwave coaxial but also has the advantage in dealing with fewer samples compared with BP neural network algorithm. The root mean square relatively error, mean absolute relatively error and maximize absolute relatively error of SVM model generalization performance are 1.06%, 0.96% and 1.16%, respectively.

[1]  Reinhard Knöchel,et al.  Resonant microwave sensors for instantaneous determination of moisture in foodstuffs , 2001 .

[2]  Stuart O. Nelson,et al.  Free-space measurement of dielectric properties of cereal grain and oilseed at microwave frequencies , 2003 .

[3]  A. Walmsley,et al.  The determination of moisture in tobacco by guided microwave spectroscopy and multivariate calibration , 2001 .

[4]  Harris Drucker,et al.  Support vector machines for spam categorization , 1999, IEEE Trans. Neural Networks.

[5]  J. Suykens.  Nonlinear modelling and support vector machines , 2001, IMTC 2001. Proceedings of the 18th IEEE Instrumentation and Measurement Technology Conference. Rediscovering Measurement in the Age of Informatics (Cat. No.01CH 37188).

[7]  F. Girosi,et al.  Nonlinear prediction of chaotic time series using support vector machines , 1997, Neural Networks for Signal Processing VII. Proceedings of the 1997 IEEE Signal Processing Society Workshop.