Research On Driving Fatigue Detection Based On EEG Signal Recognition

With the development of economy,the vehicle ownership of our country is increasing day by day.With such huge number of vehicles,it is normal for traffic accidents caused by cars.And fatigue driving is one of the main causes of traffic accidents.Brain-Computer Interface(BCI)is a humancomputer interaction way based on electroencephalogram(EEG),which can rely on the EEG signals rather than the peripheral muscles and nerves tissue to achieve free action and communication with the outside.Based on the brain-computer interface system,this paper completed the detection of fatigue driving through the preprocessing of EEG signals,feature extraction and pattern classification.The main research work of the paper has been arranged as follows:(1)A feature extraction method combining empirical mode decomposition(EMD)with energy spectrum was proposed.After the preprocessing,a series of Intrinsic Mode Functions(IMFs)were obtained by EMD method,and the IMFs were screened by mutual information to reconstruct it.Afterwards,the IMFs were extracted by energy spectrum.To verify the algorithm,we compared the features it extracted with the mainstream feature extraction methods.(2)A particle swarm optimization(PSO)algorithm optimized hierarchical extreme learning machine(H-ELM)classifier was proposed.When using multi-level perceived overrun learning machine for pattern classification,the scaling factor S and minimum error penalty ?2 affect the performance of the classifier.Therefore,this paper optimized the H-ELM parameters by PSO to enhance the classifier’s classification performance,and then extracted signal characteristics of the pattern classification.(3)The proposed method was compared experimentally to verify the effectiveness of the method.The data in experiment 1 was classified by H-ELM algorithm which was optimized by PSO and the experimental results showed that PSO-H-ELM did effectively improved the classification accuracy of EEG signals.Experiment 2 compared the traditional feature extraction method and the feature extraction algorithm of EMD combined with energy spectrum algorithm,and the extracted features were classified by using a variety of classifiers,including the PSO-H-ELM method proposed in this paper.The classification results prove that the combination methods of EMD and energy spectrum algorithm is superior to the traditional power spectrum method.