An ECG compression approach based on a segment dictionary and bezier approximations

This paper proposes a methodology for ECG (electrocardiograms) data compression based on R-R segmentation. An ECG can be seen as a quasi-periodic signal, where it is possible to find many similarities between heart beats. These similarities are explored by the proposed compression scheme through the use of a segment dictionary combined with an efficient form of progressive error codification. The dictionary is able to incorporate new patterns, in order to assure the algorithm adapts to changes in signal morphology. Experimental results reveal that high compression ratios are possible for highly regular signals, with irregular signals still achieving acceptable results.