Research On The Algorithm Of Arrhythmia Classification

With the development of social economy and the improvement of our living standard, more and more people have been suffering from cardiovascular diseases with arrhythmia the most prevalent. Arrhythmia is supposed to be the source of many fatal heart disease, and it is of great significance to detect arrhythmia for heart diseases prevention.Application of the Automatic Analysis System of Electrocardiosignal (AASE) has brought great convenience in prevention and treatment of heart diseases by shortening diagnostic time and enhancing diagnostic efficiency. This dissertation chose appropriate Automatic Detection Algorithm of Arrhythmia Classification (ADAAC) and designed the Automatic Detection and Analysis System of Arrhythmia Classification with combination of clinical experiences and computerized information technology, which has also been tested and verified by clinical diagnosis.In this dissertation, the present state and importance of this research were introduced including composition and meaning of normal electrocardiogram, inducing causes of arrhythmia and conclusion of common arrhythmia signals. After studying the preprocessing methods of ECG based on wavelet transforms, zero reconstruction of dispersion coefficients was used to filter low-frequency noise and wavelet threshold modification was used to avoid spectrum overlap, filter high-frequency noise and to prepare for automatic detection of arrhythmia. Characters of Discrete Fourier Transform, high-order statistics and QRS wave were also selected to reflect different features of ECG in frequency, statistic and morphology, based on which150features have been picked up. Finally, arrhythmia signals were classified into16kinds with successive applications of principal component analysis (PCA), principal character extraction and the KNN algorithm.Data in this research were from MIT-BIH database. Samples were divided into training group and testing group, and the Detection Algorithm of Arrhythmia Classification in this research was verified to be effective through repeated experiments.