Prediction for Gyros output based on EMD-SVR

To solve the problem of prediction for Gyros real-time output, improve the prediction accuracy of Gyros output, a new prediction algorithm based on Empirical Mode Decomposition (EMD) and Support Vector Regression (SVR) is put forward in this paper. Due to the non-linearity and non-stability of Gyros output, the conventional prediction methods are not accurate, so the Empirical Mode Decomposition were chosen to decompose the Gyros output to several Intrinsic Mode Functions (IMF) and a residual series. For these IMFs and residual series, use Support Vector Regression to build the prediction model. The prediction results of IMFs and residual series were merged to obtain the Gyros output. Result show that this prediction algorithm is effective compared with Support Vector Regression, which verified the precision of the method.