Research On Portable Mental Fatigue Detection System Based On EEG Signal

Mental fatigue has become a major problem affecting the production and life of modern people.Serious health problems and even security incidents may be caused with long-term accumulation of fatigue,which can be effectively prevented by timely detection of fatigue state.Thus,this paper studies the portable mental fatigue detection system based on single-lead EEG signals.The main contents of this paper include:(1)Research on mental fatigue feature extraction algorithm.Firstly,the wavelet packet transform was used to obtain four basic rhythm signals(?,?,?,?)of EEG,and the relative energy of the rhythm signals was used as the fatigue feature.In addition,the study found that the sample entropy of nonlinear dynamic parameters can also effectively characterize the complexity of EEG under different fatigue states.Therefore,the above two methods were combined to better reflect different fatigue states.Finally,the EEG signals of fatigue states and non-fatigue states are used to carry experiments to evaluate the performance of three feature extraction methods.Results show that fatigue assessment based on combined features is significantly better than other two fatigue assessments based on single features,which proves that the proposed method based on rhythm signal relative energy and sample entropy is a more effective method for extracting mental fatigue.(2)Research on fatigue assessment algorithm.Based on the traditional single-kernel support vector machine algorithm,a multi-kernel support vector machine algorithm with better performance is acquired by combining a Gaussian kernel function with local learning ability and a polynomial kernel function with global generalization ability.Considering that the classification effect of the multi-kernel support vector machine algorithm is strongly associated with the selection of parameters,particle swarm optimization algorithm is used to optimize the penalty factors,Gaussian kernel parameters and kernel function weight values in the multi-kernel function,to improve the performance of the classifier.Through the offline analysis of the EEG signals of fatigue and non-fatigue states,it is verified that the classification performance of the multi-kernel support vector machine after parameter optimization is significantly better than that of the multi-kernel support vector machine without parameter optimization,and also better than that of traditional single-kernel support vector machine with parameter optimization.(3)Design of mental fatigue detection system.In this paper,a portable mental fatigue detection system based on single-lead EEG signals is designed.The system consists of two parts: EEG signal acquisition front end and host computer.The front end of the EEG signal collection uses the TGAM module of Neuro Sky as the core component.The single-lead EEG signals of the subject are collected through dry electrodes,and the data is transmitted to the host computer by Bluetooth wireless transmission.The host computer detection software is designed by MATLAB,including TGAM data unpacking module,feature extraction and model training module,fatigue detection and status display module.The feature extraction in the system used the aforementioned method based on the relative energy of the rhythm signal and the sample entropy.The fatigue assessment algorithm used the aforementioned multi-kernel support vector machine algorithm based on parameter optimization.(4)Experimental research on mental fatigue detection system.An experiment with7 subjects participating in was designed and tested the mental fatigue detection system.Before using the designed system to evaluate the mental state of the subjects,the system evaluation model is trained with the EEG data of the subject collected in different fatigue states in advance.Results show that the portable mental fatigue detection system based on single-lead EEG signals designed in this paper can effectively detect the fatigue state of the subjects,and the evaluation results can be displayed and saved.The designed system is simple and convenient to access and operate,which meets the development of smart devices.