Research On EEG Signal Processing Based On Cognitive Neural Mechanism

With the development of neuroimaging technology,the EEG signal is used to study the principle that the brain produces a variety of high-level functions such as psychology,emotion,and sensation,which can improve the understanding of brains with refined structure and high performance.However,the EEG is a kind of non-stationary random signal with time-varying,strong background noise and extremely weak.Thus,precisely analyzing the hidden information of EEG is still a problem that needs to be solved urgently.In this thesis,the relationship between different task states and brain activity levels is studied firstly.Traceability analyzing the activities of the brain areas is used to raise the awareness of brain.At the same time,aiming at the problem of low recognition rate of EEG signals,a method of extracting P300 potential characteristics by combining multi-rhythm signal and CSP is proposed.Finally,from the perspective of data analysis,classifier is improved by combining the knowledge of sparse expressions to raise the classification accuracy.The main research work of this thesis is as follows:(1)Traceability analysis.Under the three music environments of quiet,flute,and zither,15 students are randomly selected as subjects to record EEG signals during the learning process.Dipole mapping is used to analyze the possible areas of the signal source in the brain and the strength of the energy,which can track changes in brain signals.At the same time,the experiment used current density reconstruction method to analyze the activity of each brain area.(2)Feature selection.For the human body in different physiological states,the energy of various rhythm signals generated in the cerebral cortex is also different.By studying multidimensional effective feature vectors of EEG signals,such as time domain,frequency domain,and spatial features,a frequency and spatial feature construction method combining multi-rhythm EEG signals with CSP is proposed in the thesis to accurately capture the effective pattern recognition feature of EEG.(3)The sparse expressions as classification.It is mainly responsible for constructing a multi-model recognition model to analyze the characteristic data.In the thesis,a classification model using sparse expressions is constructed.Among them,theelastic network not only makes the coefficient sparse,but also solves the problem of overfitting,and the generalization ability of the system is also improved.The experimental results will be explained from the following two aspects: 1)the result indicates that music may help to increase attention during the study.2)The experiment shows that the method of this thesis can obtain the effective pattern recognition feature of EEG,and the classification effect can achieve to 90.30%,which prove the high recognition rate of the current classification model.