Event detection and signal compression in digital electrocardiograms

Novel methods for event detection and signal compression are presented for the electrocardiogram (ECG). The event detection method uses a simple and fast algorithm which is highly effective. The method is based on the local and fuzzy evaluation of the size of Haar wavelet transform coefficients of the signal. The average error rate of the method was 0.68% over the entire MIT-BIH database which is of the same order as previously published results. The method achieves this level of performance without operator interference, and with globally fixed parameters which is unique in the literature. The compression method is based on active error control, and is the first method which has successfully applied a local error measure to the electrocardiogram compression problem. A resampling strategy based on the physiology of the signal is used to achieve performance improvement and computationally feasible access to singular value decomposition. The method is simultaneously capable of higher compression and higher fidelity of the reconstructed approximation than previously reported results. The average compression rates achieved by the method are 27.1:1, and 15.4:1 for the 100 and 200 series of the MIT-BIH database respectively.