An accelerometer modeling approach based on mixed-kernel support vector machine

This paper aims to model the accelerometer with nonlinearity by applying least square support vector regression (LS-SVR). A one-pass modeling approach based on LS-SVR with mixed kernel function is proposed. It could construct the linear approximated model and the nonlinear characteristics at one time. A LabVIEW-based automatic measurement system for accelerometer dynamic characteristics is designed and implemented. The results show better accuracy then conventional modeling method.