The authors present a new technique to derive from an ECG waveform the significant points, or the points where the curvature of the waveform is significant. They employ these in ECG data reduction and pattern recognition programs. The significant points are derived by first producing a chain-code from the pattern, and from that the local and then the global curvature. When only the significant points are used to reconstruct the signal, it results in a data-reduction of the order of 10:1. At the same time these significant points are used in a pattern recognition program to extract the QRS complex, and the ectopic beats.