Research On Recognition System Of Vehicle Driver’s Fatigue Status

Traffic accidents caused by fatigue driving occurred frequently, it takes great loss tothe people’s life and asset. So it is of signality and has practical application value todevelop driver fatigue early warning system. Nowadays, the study of driver fatiguedriving is quite popular. However, there are quite few products which can realize vehicle-borne, unforced and real-time processing. Existing products are affected by theenvironment, poor stability, expensive and so on problems. This paper takes advantage ofthe application of the simulator to simulate the situation of vehicle driving, designs forfatigue monitoring sensors and real-time monitoring system. Fuzzy clustering is used toprocess the multi signal features which can reflect the status of the driver fatigue.Experiments prove that the system has stronger capability of identification. It lays a solidfoundation of technology and experiment.Firstly, through analyzing research status at home and abroad and comparingadvantages and disadvantages of different methords, this paper proposes to use driver’srespiration, sphygmus, shell temperature and steering performance signals, which arevehicle-borne, unforced and easy to handle, as the criterion of fatigue. The paper designsthe hardware part of the system and prototype. The system is break up into four modules:signals acquisition, the micro-processor, power and alarm based on modularize designmethod. Especially, the sensors which are used to gather driver’s respiration and steeringperformance signals are designed.Secondly, a method for the evaluation of fatigue level which is used to determine thereal fatigue state is put forward. Datas are acquired by plenty of experiments. Throughanalysising the datas of the driver in two different state characteristics, extractingrespiratory cycle’s standard deviation, number of yawning per unit time, pluse frequency,pressure on the steering wheel, angle instability coefficient and shell temperature fordetermining the characteristic parameters of fatigue.Finally, fuzzy clustering is choosed as the core algorithms of the data processing.Established a driver fatigue model based on c-means fuzzy clustering. The results offuzzy clustering based on singal feature and multi features show that the single featurehas high recognition accuracy, the multi features has great fault tolerant capability. Thefatigue recognition algorithm is optimized by weighted average. Experiments prove thesystem has high recognition accuracy and great fault tolerant capability.