The Applied Research Of The Statistical Pattern Recognition Classification Of ECG Common Diseases

With the continuous progress and development of modern society, there is growing awareness of human beings health. Heart disease is common, frequently-occurring and also fatal issue. Before the patients arrive at hospital, physicians do not know the specific circumstances, and unable to rescue patients in time. Therefore rapid and timely diagnosis and treatment is necessary.In the thesis, first of all, the related work on electrocardiogram (ECG) automatic diagnosis is described, and 29 kinds of ECG diagnosis rules based on medical experience are summarized. The simulation shows it's more effective.After that, two sorts of statistical pattern recognition classification methods-Bayesian classifier and support vector machine (SVM) classifier are introduced, and the relative classification data making use of MIT-BIH Arrhythmia Database are got. The results of support vector machine classifier are better than that of Bayesian's. Then, support vector machine is selected to classify aiming at clinical ECG data.Finally, classifications combination approach is analyzed. The specificity and sensitivity of serial classifications combination is far from the practical application. Further analysis and improvement are needed.Now, support vector machine classifier we experienced has been transplanted to remote ECG diagnosis center. The feedback is being waited to improve the application.