Using least square support vector machine for reducing the cross-sensitivity of sensors

Classical sensors are always sensitive to several parameters in Automated Testing and Control system. This phenomenon is called cross-sensitivity. It restricts the application of sensors in engineering. In order to reduce cross-sensitivity, this paper is build a multi-sensor system measurement model. For solving the nonlinear problems in the model, multiple-input multiple-output (MIMO) least square support vector machine (LS-SVM) is used to establish the inverse model. Using the Niche Genetic Algorithm to optimize the parameters of LS-SVM, finally, it can eliminate the sensors´ output influence caused by nonobjection parameters. The result of sensors system circuit simulation model shows that: this method can get optimized parameters to suppress cross-sensitivity, and can improve the accuracy of mea surement. It is beneficial to the application of sensors.