Epilepsy Eeg Classification Based On Conditional Entropy And Sparse Representation

The diagnosis of the neurological diseases has always been the very challenging problem in the field of the biomedical science,and epilepsy is one of the most common neurological diseases,which is caused by disordered excessive or supersynchronous neuronal activities,with periodicity and repeatability.EEG plays a very important role in the diagnosis of the epilepsy.Traditional clinical methods mainly rely on the EEG for visual detection,which is time-consuming and laborious,and may produce unnecessary human experience error.Therefore,it is necessary and meaningful to classify the epileptic EEG signals automatically.The brain electrical signal contains a variety of physiological and pathological information,and provides the basis for the diagnosis and treatment of the neuropathic brain.Considering the height of the brain electric signal complexity and nonlinear characteristics,this article from the two angles to study the classification of the epileptic EEG signals,that is,nonlinear feature extraction strategy and sparse representation method,and the main work is as follows:Firstly,the nonlinear feature extraction is mainly based on the changes of some brain activity characteristics brought about by the changes of the brain state to analyze and detect different brain states.As a nonlinear measurement method,conditional entropy can effectively measure the irregularity of the physiological signals.Conditional entropy is used to analyze the complexity of the EEG signals,and based on the neural mass model and the variation coefficient,the parameters and performance of the conditional entropy are selected and analyzed,and the automatic classification of the epileptic EEG signals is realized by combining the support vector machine classifier.Secondly,the sparse representation method is to compare the reconstruction errors of the training samples under different categories to the testing samples through the sparse constraint solution of the₁l norm minimization problem,so as to realize the effective classification of the epileptic EEG signals.Different from the traditional base pursuit de-noising method to solve the equation constraint problem by minimizing the₁l norm of the unknown parameters,this paper uses the fast iterative shrinkage threshold algorithm to solve the problem to determine the category of the epileptic EEG signals,and determine the number of the dictionary atoms and the reconstruction error tolerance with the best classification effect.The EEG classification of the epilepsy signals based on conditional entropy and sparse representation provides a basis for the diagnosis of the epilepsy patients,and could slow down the work intensity of the medical staff and improve the work efficiency.