The Study Of Driving Fatigue State Recognition Based On Pulse And Eye Movement Features

With the development of social economy,the number of motor vehicles in China is growing rapidly,and the accidents is increasing.Fatigue driving,as one of the important inducing factors of traffic accidents,poses a great threat to people’s lives and property safety.Therefore,the detection of fatigue driving status has great economic and social value.There are three main kinds of fatigue detection methods:physiological signal based detection,face feature based detection and vehicle condition based detection.Physiological signal detection is usually based on EEG,ECG and other signals.The signal accuracy is high,but the sensor can easily affect the driver’s operation.The detection based on facial features is more convenient and will not affect the driver’s operation,but it is easy to be disturbed by background noise and has a weak stability.The detection based on car status is usually based on wheel rotation.The detection of vehicle trajectory has simple structure and good real-time performance,but the anti-jamming ability is insufficient.Driver’s driving mode will have a great impact on the final result judgment.In this paper,a driving fatigue state detection and recognition model based on pulse and eye movement characteristics is proposed,which combines the physiological signal processing technology and image processing technology with machine learning,aiming at the problems of poor detection stability and weak anti-interference ability of the single signal source mentioned above.The time domain and frequency domain features are extracted from the pulse signals,and the features are filtered based on the extreme random tree.Eye movement features are extracted from image information.Eye movement features are extracted based on multi-level threshold construction.A classifier is constructed using Adaboost algorithm to detect and recognize driving fatigue.The main contents of this paper are as follows:In order to ensure the safety of the experimental process,a more realistic simulated driving environment was built.The subjects simulated the real driving state in the simulated driving environment.The subjects collected the pulse signals and facial image signals of the subjects in awake state and fatigue state respectively.Using reliable data acquisition equipment,healthy subjects without major medical history were selected,and effective fatigue stimulation method was used to carry out experiments in an experimental environment independent of the signal acquisition part and the experimental detection part.Pulse signal is preprocessed based on wavelet transform.The noise and baseline drift of the pulse signal are removed based on Sym8 wavelet.According to the pulse signal,multiple time series,such as the time span of a single pulse cycle and the area under the pulse wave of a single cycle,are obtained.Mean value,mean square error,high frequency,low frequency and other time and frequency characteristics of these time series are calculated.Face alignment based on ensemble regression tree is used to obtain face regions,and eye feature points are indexed.Eye movement features are extracted from human eye regions based on multi-level thresholds.Using image segmentation,image scaling,gray scale,binarization,pixel color detection and other image processing methods,we extract a variety of eye movement features,including the upper eyelid height,eye area,the proportion of pixel color in a specific area,and so on.Feature screening and fatigue detection and recognition are carried out based on extreme random tree and Addaboost.The extracted pulse features are filtered by using extreme random tree to get key features,and the sample set is constructed by using key features.The classifier is constructed based on AdaBoost algorithm,and the fatigue recognition monitoring model is constructed.The fatigue state is detected and recognized by using pulse and eye movement features.