Research On Fatigue Identification Method Based On Characteristics Of Driver Behavior And Eye Movement

With the rapid development of road transportation, rapidly increasing vehicle population, traffic accidents, especially the serious traffic accidents occurred more and more frequently. Researches have shown that fatigue is one of the leading causes of road traffic accidents. Therefore, the study of driving fatigue identification system, choosing the reliable and efficient detection indexes, builtting testing model is of great significance on the improvement of road traffic safety and reducing the frequency of traffic accidents.Based on the analysis of the driver’s eye movement and the driving performance under different fatigue, extracting the driving performance and eye movement characteristic parameter. This paper develops driving fatigue identification model in use of information fusion theory, achieving the higher identification accuracy. The specific research contents are as follows:(1) Experiment program design and data collection. Driving performance data and eye movements data from 10 drivers were collected synchronized based on driving simulation. Fatigue level was scored on the KSS standard and driving performance data and eye movements data were coded and catalogued according to the KSS score, training and test database for different fatigue level was established.(2) The paper studies the change rules and characteristics of driver’s driving performance and eye movement in the different fatigue state in the use of statistical methods, especially analyzes the steering wheel angle, steering wheel angle speed, speed and acceleration of the vehicle; in aspect of eye movement, the paper analyzes the features of blink, fixation, saccade, and eye movement characteristics of pupil diameter changes in the different fatigue state. Under different time window to extract the characteristic parameters, the different level of every characteristic parameters were tested by the analysis of variance (ANOVA), with varying fatigue levels. Finally, the extraction driving performance characteristic parameters include:SAM, SASTD, SWM, SWSTD, Am, Vstd and Astd; eye movement characteristic parameters include:BF, FIXT_mean FIXT_std, SACV_std and CVPLD.(3) This paper generalizes the common methods of information fusion, analyzes the advantages and disadvantages of the fusion method, determines the BP neural network as a fatigue identification fusion algorithm of the model, then discusses the fatigue recognition model based on BP neural network in detail, the establishment of the process including the selection of learning algorithm, the number of input and output neurons, the number of hidden layer of network, the number of hidden layer neurons, uses neural network toolbox of Matlab software function program. The training sample and test sample were randomly selected for training and validation of the model. The results show that the average identification accuracy rate of the driving fatigue identification model are up to 83%、69.6% and 79.6% on the different fatigue state. So that the model can be used to identify the driver fatigue states.