Study On Detection Of Fatigue Driving Based On Steering Wheel Grip

In china, the number of traffic accidents decreased in the past few years, but the deaths caused by the accidents has been increasing year by year. Fatigue driving was contributed to the traffic accidents, and fatigue driving is a key research topic. By analyzing the research status of fatigue detecting at home and abroad and comparing the characteristics of different fatigue detecting methods, the fatigue detecting methods were mainly divided into three categories: detecting the drivers’ face character by computer vision, detecting the drivers’ physiological signals by using the physiological equipment, or detecting vehicle information by using car equipment. Compared with the other fatigue detecting methods, this paper takes advantage of steering wheel grip signal to detect driver fatigue, which has the following advantages, like easy signal acquisition, high signal noise ratio, low cost, and not affect the normal driving operation. However, the detecting accuracy of this method was relatively low, and the use of EEG to detect driver fatigue shows highly accuracy, which only widely used in the experiment. So this paper simultaneously detected the drivers’ EEG signal and grip strength signal, and then used the BP neural network to establish the relationship of fatigue characteristic parameters based on two different signals.Firstly, the experiment was conducted to monitor the EEG and grip strength signal, and then six drivers respectively performed 30 minutes normal driving and fatigue driving for the sake of drivers’ safety. Secondly, on the one hand, this paper preprocessed the raw EEG signal by setting 10 s time window, then extracted the time-domain fatigue feature by wavelet packet analysis method, and computed the average power to reflect EEG fatigue index R, and determined the drivers’ fatigue state. On the other hand, in terms of grip strength signal, the time-domain feature and the frequency-domain feature were extracted and computed the average in the same time window. Using the independent sample t-test analysis to screen characteristic parameters of significant differences exist between the normal driving and fatigue driving, the grip strength characteristic parameters were smoothed to represent the fatigue state. Finally, BP neural network model was used to predict the driver fatigue state, taking the steering wheel grip strength as input layer and R values as output layer, and this paper trained the prediction model and verified its accuracy.