Research On Non-Contact Fatigue Driving Detection Method Based On CEEMD And Extreme Learning Machine

With the improvement of people's living standards and the development of automobile industry,the number of cars on the road has continued to rise,and numerous accidents occur every year.Many studies have shown that fatigue driving is one of the main causes of traffic accidents and hidden dangers of road safety.Therefore,it is significant to study the real-time online detection method of driver state degree to ensure traffic safety.This thesis designs a non-contact fatigue driving detection system based on complementary ensemble empirical mode decomposition(CEEMD)and neural network.Firstly,the hardware system is designed based on the heartbeat and respiratory signals contained in the micro-Doppler signal.After the initial signal is processed by impedance matching,differential amplification,active band-pass filtering,level shifting and analog-to-digital conversion,the digital radar signal with a passband of 0.1-5Hz is generated.Then according to the characteristics of radar signal,the traditional FIR and IIR digital filters and CEEMD algorithm based on complementary sets are used to decompose the signals.The advantages and disadvantages of the three methods are compared.In view of the large adjacent frequency noise and phase distortion of the former two,which is not conducive to the extraction of signal features,this thesis adopts the CEEMD algorithm,which greatly retains the time-domain and frequency-domain features of the initial signal about heartbeat and respiration.Finally,the neural network is used to take the decomposed signal as the feature input,and the fatigue state based on the expert evaluation method of facial video is used as the output to train the data set.In this thesis,two algorithms,back propagation(BP)neural network and extreme learning machine(ELM),are studied.Due to the limitations of traditional neural networks,the thesis uses the momentum optimization BP neural network(MOBPNN)and the differential evolution extreme learning machine(DEELM)to distinguish the fatigue level.In this thesis,a micro-Doppler signal acquisition platform is built,and the feasibility of the system is verified through simulated driving experiments.The experimental results show that the built system can effectively detect the fatigue state of the experimenters.The best detection accuracy of DEELM is 82.1%.Compared with the MOBPNN and the classic ELM,it has better performance in fatigue level detection.