A framework for ECG morphology features recognition

ECG morphology features, as a kind of significant diagnosis feature, are widely used by experienced cardiologists and highlighted in professional medical textbooks. Fail to utilize it should be one of the most important reasons for the underperformance of automatic ECG classification. In this paper, a framework for ECG morphology features recognition is presented. 1-nearest-neighbor with dynamic time warping (1NN-DTW), a strong time series matching algorithm, is used to compare the ECG segments with the templates store in the system. Template selection and reduction is applied to accelerate the classifier and cut down the templates volume as well. With the help of new proposed template reduction algorithm, our system has an accuracy of 90.71% by using a small portion of the original template set.