An Study And Application Of Pathological Signal Analysis Method Based On Change Point Detection

With the development of computer science and technology,we have entered the era of "data explosion",the daily generation of huge amounts of data in various industries continue to accumulate,but also formed the intangible assets of enterprises and society.Large amounts of data contain potentially a wide variety of valuable information,which is increasingly valued by all sectors of society,and the application of data mining technology is born.Due to the large fluid volume,strong continuity and different modalities of time-series data,how to quickly analyze it and mine valuable information has become a research hot spot.As a branch of data mining,the data stream mutation point detection technology can filter out most of the normal data from the overall data distribution,quickly locate the mutation point location,and analyze the cause of mutation and data fluctuation before and after.Among them,the sliding window model can segment the data flow to form several sub-windows for separate analysis,effectively improving the data processing capability and mutation point detection rate.The performance of the mutation point detection algorithm,the size of the sliding window and how it is updated are all important factors that affect data mining.Neurological diseases,such as epilepsy and heart disease,are characterized by high morbidity,disability and mortality,and are a major threat to human life and safety.In this paper,we propose the theory of establishing fluctuation vectors for time-series data based on mutation point detection algorithm and sliding window model,and give two template matching algorithms based on fluctuation vectors,which can realize the fast analysis of large amount of time-series data.And selected epileptic brain electrical signal data for applied analysis,verified the performance of the algorithm and achieved significant results.First,this paper introduces a better performing TSTKS mutation point detection algorithm.The lack of accuracy and time-consuming detection of common mutation point methods was effectively addressed.In order to solve the disadvantage of not being able to detect multiple mutation points and to improve the detection speed,the sliding window theory was introduced and a mutation point detection model based on the TSTKS algorithm and sliding window was proposed.The results of multiple comparative experiments show that the model has a better detection of multiple mutation points for timing data compared to other algorithms.The optimal window size was then determined by exploring the effect of sliding window W variations on detection performance at specific data lengths.Secondly,the theory of building time-series data fluctuation vector based on mutation point detection and sliding window is proposed and the template matching algorithm based on time-series data fluctuation vector is given.Using the TSTKS mutation point detection algorithm and the sliding window model,the fluctuations of each window are calculated,and the fluctuation vectors of the data are integrated to characterize the fluctuations of the data,and two matching classification algorithms based on template vectors are proposed,respectively.The simulation results showed that the template matching algorithm based on correlation coefficients is suitable for the rapid detection analysis of interictal data,and the algorithm based on statistical fluctuations and SVM has better classification effect on interictal and preictal data.The two algorithms have different application scenarios and both validate the feasibility of analyzing fluctuation vectors as time-series data features.Finally,in order to solve the problems of complex feature extraction,high computational volume and insufficient real-time during epilepsy pathology signal detection,a customized pathology expert system based on fluctuation vectors and SVM is proposed.Simulation of the patient's electroencephalopathological data for real-time detection of early warning,the simulation results showed that the expert system is fast and sensitive to the diagnosis of epilepsy disease,and will give warning signals at least 1100 s before the onset.Although there are some false alarm signals for some patients,overall,it can provide a certain amount of treatment and preparation time for epilepsy patients,meet the needs of the actual epilepsy warning,and promote the application of epilepsy automatic detection technology in clinical diagnosis.