Skip to Main Content
ECG data compression algorithms are important for storage, transmission and analysis. An essential requirement of the compression algorithms is that the significant morphological features of the signal should not be lost upon reconstruction. In this paper two different neural network based methods are investigated for ECG data compression. The first method uses filters for attenuating noise and interferences, a radial-basis function network for the detection of R-points for separating the waveform into different cycles and finally multilayer back propagation networks for data compression. In the second method, the back propagation networks are used as nonlinear predictors for achieving the data compression. Compression results obtained by using the two different methods are evaluated based on standard MIT-BIH ECG Test Database.