Driver’s Mental Status Assessment Based On Fusion Of Visual Information And EEG

In recent years,with the rapid development of the national economy,the number of families with cars is increasing,and the number of motor vehicles on the road is increasing.Frequent traffic accidents have become the main problems facing the country today.In all the major traffic accidents,fatigue driving is one of the most important factors leading to the accident,so the study of the driver’s mental state assessment method has become a major issue.In this thesis,a simulated driving test platform is set up in the laboratory,and the experimental data collection specification is designed strictly.At the same time,the face images and EEGs are collected by visual sensor and Emotiv Epoc device.The database samples of this experiment are set up and studied in the following aspects.For visual information,we propose Adaboost algorithm and Haar feature to detect face images,generate convolution neural network sample library,design the convolution neural network identification framework,and analyze its structure and parameter optimization.The multi-scale network model is used to extract the local features and global features,and the dimensionality of the network is reduced by designing a 1×1 convolution kernel to improve the recognition accuracy and training time.Compared with the traditional manual extraction method and the traditional convolution neural network model,the experiments show that the proposed network training has less energy loss and higher recognition accuracy.According to the EEG information,First of all,through the simulated driving system and Emotiv Epoc brain cap to collect human EEG.Using Butterworth filter to remove high-frequency interference,HHT transform to remove brain electrical noise,the use of ICA algorithm to eliminate eye interference Through the discrete wavelet transform and wavelet reconstruction to analyze the rhythm of wave energy,extract the driver’s mental state EEG features,input to the SVM classifier,the classification model to identify the driver’s mental state.The experimental results show that the EEG is a gold standard to measure the behavior of human body,which can identify the driver’s mental state very well.In this thesis,multi-source signal fusion methods are analyzed to determine the fusion in the decision-making level.The fusion of visual signal and EEG is selected based on DS evidence theory.According to the task characteristics of this thesis,the probability distribution function and synthesis rule are improved,and the visual and EEG fusion classifiers are established.The experimental results show that the classifier constructed by information fusion method is better than single information classifier.