Study Of Driving Fatigue Recognition Algorimm Based On Eye-movement And Pulse Information

Traffic safety, as a hot topic, directly relates to people’s livelihood. The driving fatigue, as one of the major problems causing traffic accidents, has received the widespread attention. Driving fatigue may reduce the drivers’ attention and reaction ability, which is easy to cause traffic accidents. According to statistics, traffic accidents caused by driving fatigue accounts for 20% of the total number of accidents, 40% of the heavy traffic accidents, and 80% of the traffic mortality. Thus, the driver fatigue has become the main hidden danger of traffic accidents, and it’s hardly to detect and monitor. For this, a large number of scholars have researched on driving fatigue, and achieved some results.In order to solve the problems in the process of driving fatigue recognition, like the low identification accuracy and bad practical performance, a driving fatigue recognition algorithm based on fusing of eye movement and pulse information was proposed in this thesis. The algorithm extracts the fatigue characteristics from eye movement and pulse information and using extreme learning machine (ELM) as the classifier, to identify whether the driver is in the driving fatigue state or not. The major jobs of this thesis are the following:1) Analyzed and summarized the forming reason, social harm, research status and existing problems in current research of driving fatigue.2) Eyes movement information processing. Analyzed the theory of mixed integral projection algorithm and mean-shift algorithm, and then a feature extraction method base on these two algorithms was proposed. This method includes four steps: background removing, eye location, eye segmentation and feature extraction. It can consider the location and luminance information of the eye together, and accurately extract three eye fatigue characteristics:PERCLOS, blinking frequency and average closing time.3) Pulse information processing. Similar to the fatigue characteristics of ECG, three characteristics of pulse:the mean of main wave interval, the standard deviation of main wave interval and the high-frequency to low-frequency power ratio were chosen as driving fatigue recognition features. To recognize driving fatigue by pulse information was proved to be reasonable.4) A driving fatigue recognition algorithm based on ELM was proposed. By fusing of eye movement and pulse information, this algorithm can identify whether the driver is fatigue or not. Then, a simulative fatigue driving experiment was given to verify the performance of this algorithm. The experimental result shows that, this algorithm can further improve the recognition rate of fatigue driving.