Design And Application Of Electrocardiogram Diagnosis System Based On Multifractal Theory

In the twenty-first century,the population of the greybeard has been increasing and the existing medical service resources are not evenly distributed in urban and rural areas.The concern of residents' health problems in China is gradually being strengthened.Cardiovascular and cerebrovascular is one of important diseases endangering the health and life of modern people,and arrhythmia is an important manifestation of cardiovascular and cerebrovascular diseases.Accurate classification and identification of ECG is a prerequisite and key to diagnosis of cardiovascular and cerebrovascular diseases.However,it is unrealistic for physicians with rich clinical diagnosis experience to continually carry out cumbersome,boring and repetitive ECG recognition.So in order to keep the doctor with rich experience from the cumbersome,easy-to-fatigue graphic recognition work,and let them pay more focus on the patient's symptoms and disease diagnosis and treatment.In this paper,an ECG automatic classification and recognition algorithm based on multi-fractal and neural network is studied.This study will improve the speed of application programming of medical automatic diagnosis and early warning of major diseases in China,and it is of great practical significance to further improve the level of medical automation in China.At this stage there are many ECG automatic diagnosis and identification systems,but they generally have a common shortcoming,that is their research direction more inclined to time domain and frequency domain analysis.Because the heart is a complex nonlinear chaotic system which is affected by many factors,the form of ECG is different.Time domain and frequency domain analysis can only be given the time-domain and frequency-domain characteristics of the data,and with the external factors and the internal together working,there is a great gap between the original expected results and the actual situation.As an important research content of nonlinear chaotic systems,fractal is gradually used to solve the problem of chaotic systems.A large number of researchers have proved that the ECG signal has multi-fractal characteristics,while the use of multi-fractal to analyze the chaotic system is also a trend.In this paper,the main research content is ECG automatic identification research: a)an algorithm of automatic segmentation of ECG signal using differential threshold method is designed and realized.The algorithm can automatically identify each ECG cycle in a segment of ECG signal and ignore incomplete ECG cycles;b)an algorithm that uses multifractal theory to describe data features to realize data classification is proposed.The multi-fractal and semispectral characteristics of ECG and generalized Hurst exponent are used to train and test the neural network model.The ac curacy of model arrive to 97%;c)a complete ECG signal annotation system is constructed,which c an automatically identify a segment of ECG with multiple cycles and annotate each cycle.At the same time it can automatically ignore the incomplete ECG cycles at the beginning and end of ECG signal,so it has a better fault tolerance.