Driving Style Recognition in the Vehicle Starting Condition Based on the Modified Fisher Discrimination

According as the driving data of the vehicle starting condition collected from experiments, the driving style characteristic parameters were selected based on principal component analysis (PCA) method. Then, clustering and analyzing the driving data with Gaussian mixture mode (GMM) clustering algorithm, a driving style recognizer was developed based on Fisher discrimination. Finally, the classical Fisher discrimination and the modified Fisher discrimination were utilized to identify the test set of driving style data comparatively. The results show that, the recognition accuracy with modified Fisher discrimination can reach more than 85%, proving the availability and veracity of this modified Fisher discrimination in the estimation of driving style and vehicle moving performance.