Research On Fatigue Driving State Recognition Method Based On Functional Data Analysis And Kernel Principal Component Analysis

Fatigue driving is one of the main factors causing traffic accidents at present.As a physiological indicator that accurately reflects brain activity,EEG signals have un-paralleled advantages in the detection of working conditions of personnel,it is a"golden method" for analyzing and judging whether the human body is in fatigue state.EEG signals are non-stationary,continuous and non-periodic,it is of great significance and application value to study the effective method of fatigue driving state recognition based on EEG signals under these characteristics.The main work and innovation are as follows:(1)A method of EEG description and feature extraction based on functional data analysis is proposed.In order to make the extracted EEG signals features better reflect the continuity,internal dynamic changes of the signals,and the need to mine more abundant data information,a method of EEG signals description and feature extraction based on functional data analysis is proposed.Firstly,the collected discrete EEG data are functionalized to describe the real-time continuity of EEG information.Then,by extracting the EEG signals at the extreme value of the curve between the normal and fatigue states as the feature,it is intended to be used for fatigue driving state detection.(2)A fatigue driving state recognition model based on functional data analysis and KPCA is proposed.Firstly,aiming at the nonlinearity of EEG signals and the problem that the feature dimension extracted based on functional data analysis is too high,KPCA method is used to reduce the dimension of the extracted feature.Then,the recog-nition model of fatigue driving state is constructed by selecting a suitable classifier.Finally,the test is carried out on the data set collected by the simulated driver,the ex-perimental results show that the proposed method based on functional data analysis and KPCA has a good recognition rate.(3)The stability and application convenience of the method are analyzed.The number of electrodes used in current recognition of fatigue driving state based on EEG signals is generally whole electrode,which is difficult and inconvenient in practical application.Therefore,two kinds of electrodes are used to test the model,one is to test the validity of the extracted features and the stability of the method,the other is to study the convenience and feasibility of practical application.The EEG data of the whole electrode and the combined electrode were tested,and the experimental results show that the proposed method has good recognition effect and stability.