We propose an input delay neural network (IDNN) based time series prediction algorithm for compressing electrocardiogram (ECG) signals. Our algorithm has been tested and successfully compared vis-à-vis other popular techniques for its compression efficiency and reconstruction capability.