NEW TECHNIQUE FOR ECG DATA REDUCTION AND PATTERN RECOGNITION.

A new technique is described to extract significant points from electrocardiogram (ECG) waveforms. This algorithm extracts significant points representing contours and inflections in a waveform. The curvature at a point is calculated from the degree of unexpectedness, the scan length, and the scan tolerance. The significant points are derived by a two-pass algorithm that determines the forward and the backward end-points of a contour and merges the points of low significance. The resulting array of significant points, as compared with the original data array, represents a data reduction in the range of 1:5 to 1:10. The authors are able to extract the QRS complex and the P and the T waves by setting appropriate scan interval and tolerance.