On system identification of laser gyroscope static drift

Laser gyroscope as a kind of photoelectric sensor, the static drift output data generally validate to be a non-stationary time series, and traditional Time Series Identification lacks dependable modeling precision. Analyzed the advantages of RBF Neural Network and SVM (Support Vector Machine), combined Wavelet Transform with ANN and SVM, studied the identification methods of WT-RBF (Wavelet Transform RBF Neural Network) and WT-SVM (Wavelet Transform Support Machine), and respectively constructed four kinds of model: RBF, SVM, WT-RBF and WT-SVM. Compared the modeling precision using six performance indexes, the results of computer aided analysis indicates that combined forecasting method can enhance the precision greatly than any other single methods in meeting the performance indexes.