Long-time EEG Detection Baesd On One Class Support Vector Machine For Epilepticseizure Time

Epilepsy has clinical manifestations and EEG signals.Generally,the two are combined to judge,but some clinical manifestations are not obvious.At this time,EEG signals are mainly used to judge.Therefore,EEG signals are an important basis for judging epileptic patients.However,the EEG signal is obtained by placing appropriate electrodes on the patient's scalp.It is conceivable that there is a very large artifact in the scalp EEG signal.Seizures are very random and it is often difficult to record seizures,so to get better results,you need to record a patient's complete wake-sleep-wake cycle,and the monitoring time ranges from hours to days.At present,processing a large amount of data and analysis all require manual processing by medical workers,which is not only a huge workload,but also difficult to feedback to the clinic in real time.It may also lead to the same data,but the results of different scholars analyzing the data are different,and there is a very lack of correct and time-saving methods.In this paper,the study of seizure time based on support vector machine for long-range EEG detection is carried out,and the following research works are carried out:(1)Obtaining data and preprocessing it: According to the data of real epilepsy patients and normal people,read in medical proprietary format,calculate the binaural leads,preprocess the data,remove errors and unrecorded Data,remove artifacts,data redundancy analysis,data standardization,and finally obtain higher quality data,which is convenient for subsequent research.(2)Based on the Recurrence Time method,feature extraction can be performed on longrange EEG data,and for a large number of long-term EEG data,the results can be quickly obtained,and the scalp with large artifacts can be processed.EEG signals,from which thecharacteristic variables of seizure signals can be accurately found.(3)Support vector machine method based on single classification is a method for detecting abnormal data.It can classify imbalanced medical data better.Through training,constantly adjusting the model parameters,to obtain a good model,the seizures and normal types can be separated,and the accuracy is improved.(4)Optimize the experimental data,use power spectral density to distinguish the false difference and seizures;analyze the experimental data results,and compare and analyze with the SDLE method,and find that the accuracy is better.This article deals with the analysis of digital EEG signals and their clinical application.It mainly includes analyzing non-linear methods based on complexity science recursive time,analyzing EEG signals,extracting EEG features,and then identifying abnormal signals from their features,assisting doctors in determining seizures,reducing doctors' working hours,and reducing epilepsy.Issues of early warning.Project research has important research value and scientific significance.