Early Forecast and Recognition of the Driver Emergency Braking Behavior

The driver emergency braking behavior to be distinguished and predicted exactly was difficult. In order to gain the testing data of driver emergency braking action, 7 professional drivers were selected and 3 scenes of driver braking behavior were designed and simulated by means of road test. And the testing data were captured by the data acquisition system with sensors. Utilizing relative fuzzy membership degrees, the testing data were normalized for the probability neural network (PNN). Under different number of training sample data selected from test data, neural network construction model based on the PNN was built and simulated. Results show that when the number of testing sample data is 260 the hit rate is 95.3%. And more, the results indicate the validity of fuzzy normalization and PNN with adequate road testing data, consequently, are an effective method for recognition and prediction of the driver emergency braking behavior.