Brain Fatigue State Research Based On EEG Signal

Electroencephalo-gram(EEG)is an electrical signal produced by neuronal activity in the brain.EEGcan be acquired by the corresponding equipment,and it is an important method for the study of brain function and the diagnosis of brain diseases.EEG signal can timely reflect the state of brain activity,including human brain fatigue.Analysis and monitoring of human brain fatigue by EEG signal analysis is a challenging task in the field of signal analysis and biomedicine,involving many emerging research methods.Fatigue can not directly endanger human life,but long-term fatigue can harm human health.In the operation due to fatigue caused by misuse can easily lead to security incidents.Fatigue has become a threat to human in daily production.Now we need an effective method which can monitored the degree of fatigue,so that to protect human life.Therefore,how to quickly and accurately monitor the human body fatigue state is very important.EEG monitoring and evaluation of human brain involved in EEG signal processing multiple levels,the paper research workingin the following areas.In the EEG signal denoising: the traditional wavelet soft-threshold function will lose some of the details of the features,the wavelet hard-threshold function will produce pseudo-Gibbs effect.During recent years,some new wavelet threshold functions was improved.These functions only aimd for a particular signal has a better effective,but can not be self-adaptive for any signal.Aiming at these problems,this paper presents a method of EEG signal denoising based on nonlinear continuous attenuation of wavelet adaptive coefficient.Firstly,the wavelet coefficients are decomposed by the appropriate wavelet basis function,and then the wavelet coefficients are processed by using the nonlinear continuous attenuation function.Finally,the wavelet coefficients are reconstructed.The nonlinear continuous attenuation function of wavelet coefficients constructed in this paper can adapt to the denoising effective in the system,andfunction can adaptively adjust the SNR and RMSE to achieve the best denoising effect.Experimental results show that the nonlinear continuous attenuation function can effectively avoid the details loss and the pseudo-Gibbs problem caused by artificially setting the threshold function in the traditional wavelet denoising process,and the SNR and RMSE after denoising are both excellent than the traditional wavelet denoising method.In the aspect of EEG feature extraction,the wavelet packet transform and the Hilbert-Huang transform are used to improve the feature extraction process in order to solve the problem of frequency overlap in the feature extraction of traditional wavelet transform.Based on the analysis of the sampled signal,the stationary wavelet packet transform is used to extract the high frequency and low frequency sub-band of the EEG signal,and the extracted subband signal is decomposed by EMD,so that the nonstationary EEG signal is transformed into linear steady state signal.Hilbert spectral analysis method calculate the instantaneous sub-band signal frequency and instantaneous amplitude.The extracted instantaneous frequency and instantaneous amplitude are combined into eigenvectors for classification.Based on the classification of EEG signals,this paper combines PSO with least squares support vector machine(LS-SVM).Compared with SVM,LS-SVM suitable for a large number of data processing and improves data processing speed.The selection of the penalty factor C and the kernel parameter g in the kernel function is quicker than other intelligent search algorithms such as genetic algorithm and particle swarm optimization,so that the parameters can be quickly optimized in the training process of the classifier.In this paper,the EEG data of 0,8,16 and 24 hours of 24-hour fatigue experiment are selected,and the instantaneous frequency and instantaneous energy extracted by Hilbert-Huang transform are combined as eigenvectors.To the least squares support vector machine classifier based on particle swarm optimization,the accuracy of data analysis is 85.83%.Compared with the general support vector machine(SVM)classification algorithm,the accuracy of the classifier using the PSO LS-SVM is greatly improved.In this paper,the design and implementation of online real-time monitoring of brain fatigue status analysis software.The software is based on the Net Framework 4.5 platform Visual C # development of the visual interface program can be real-time access to the EEG signal online analysis,and signal window length adjustable,according to the experimental requirements to intercept the required signal length to achieve the API function forthird-party software or other researchers to carry out the development and verification.The results of online fatigue monitoring and grading software prove that the accuracy of the fatigue analysis is close to that of offline data analysis.