Research On Multi-view Clustering Model For Epileptic EEG Signals

With the continuous development of artificial intelligence technology,the integration of the traditional medical and health system and artificial intelligence technology has formed a smart medical treatment that is now deeply rooted in the hearts of the people.Smart medicine has the characteristics of rapid and accurate diagnosis of the condition,and plays an important role in the diagnosis and treatment of common brain diseases like epilepsy.Epilepsy is caused by the abnormal discharge of the patient’s brain,which is manifested in the patient’s EEG as sharp waves,slow waves,spike waves,and other characteristic waveforms of epilepsy.Smart medical uses advanced technologies such as signal recognition and machine learning to identify and analyze the biological signals fed back from the subjects’ brain electrical signals,and provide diagnostic results.This is the main method for smart medical treatment of biological signals.In the past,doctors used their own experience and theoretical knowledge to judge whether there are characteristic signals by observing the subject’s EEG signal to realize the judgment of the condition.This method of diagnosis through observation often infuses the doctor’s own subjective judgment,leading to misdiagnosis of the condition and low diagnosis and treatment efficiency.Therefore,it is of great significance to realize the automatic detection,extraction and recognition of the characteristic waves of the EEG signal.With the continuous development of advanced technologies such as artificial intelligence and signal recognition,this provides new ideas for the realization of EEG signal recognition and processing technology and opens up new development paths.This article is based on epilepsy EEG signal data,realizes EEG signal processing and uses machine learning methods to realize EEG signal identification and diagnosis.The specific work is as follows:1)Judging the waveform data of epilepsy EEG signals mainly uses machine learning methods such as classification and clustering to realize epilepsy diagnosis.The traditional machine learning diagnosis method is to directly classify or cluster the characteristic data of epilepsy EEG signals,and use the results as the basis for diagnosis.Because EEG signals can be divided into multiple perspectives according to different feature extraction methods,the above method ignores that each perspective has different pathological diagnostic value.Therefore,there is no significant improvement in the accuracy of the final disease diagnosis.For this reason,this article adopts a multi-view strategy to analyze EEG signals of epilepsy to improve the accuracy of diagnosis.2)Because EEG signals have the characteristics of high randomness,small amplitude,susceptibility to interference,and non-stationarity,this paper realizes the processing of epileptic EEG signals from multiple aspects based on the existing processing of biological signals.The first is to reduce the noise of the EEG signal.Here,the independent component analysis algorithm and the wavelet transform algorithm are mainly used to remove the interference of other signal pulses from the EEG signal including ocular artifacts.Then,wavelet transform,short-time Fourier transform and nuclear principal component analysis are used to filter the denoised EEG wave,highlight the corresponding frequency band characteristics in the signal,and amplify its characteristic waveform to realize the extraction of the characteristic wave.3)For the diagnosis of epilepsy EEG signal data,classification and clustering can be used to achieve the ultimate goal.However,due to the huge amount of EEG data,the classification method requires a lot of time and cost to complete each piece of data.Labeling is not advisable in actual medical diagnosis.Therefore,this article combines the characteristics of multiple perspectives of EEG signals and the huge amount of data,and uses multi-perspective clustering analysis to realize the identification and diagnosis of epileptic EEG signals.When traditional multi-view clustering methods process multi-view feature data similar to EEG signals,their common practice is to treat each view as a separate sample,and perform a separate cluster analysis on each view sample.After obtaining the individual clustering analysis results of each perspective,the integrated learning strategy is used to integrate the individual clustering results of these perspectives,and the integrated result is the final result obtained.In this way,the multi-view data processing strategy that separates each view independently for analysis is very likely because the clustering results under a certain view have obvious deviations,resulting in inaccurate learning results or unstable algorithm performance.To this end,this paper proposes a weighted double exponential multi-view clustering algorithm(DIC-MV-FCM).Its multi-view weighted adaptive strategy can effectively control the importance of each view,and realize that each view can be clustered in the process of clustering.Effectively adjust the weight relationship to make the weight of the perspective with the best clustering effect and performance as large as possible,and improve the performance of clustering.4)The experiment of the weighted double exponential multi-view clustering algorithm proposed in this paper shows that it has a good diagnosis effect of epilepsy and reduces the time consumption of diagnosis and treatment.At the same time,according to the importance of different perspectives,corresponding weight adjustments are made to more adapt to the actual diagnosis and treatment of epilepsy.