Abstract: For achieving a higher compression ratio (CR) in compression sensing, the time-sparse bio-signals, such as electrocardiograph (ECG), are generally directly filtered via a dynamic or fixed threshold, however, inevitably leading to the loss of critical diagnostic bio-information. We propose a compression scheme to reduce the transmitting loss. Instead of the directly utilizing the original ECG data, the residuals between original and synthetic ECG signals are applied as the input signal. We employ the dynamic model to guarantee the consistency between the synthetic ECG signals waves (P, Q, R, S, and T) and the originals. The feasibility of the proposed method is tested through operating on the recorded ECG signals from a healthy human. During the process of building simulation platform, the sparsity, percentage root mean difference (PRD) versus sampling frequency, and signals reconstruction algorithm are fully taken into account. Before compression, we set the threshold to filter the residual waves, in which utilizing the residuals as input data by setting the thresold as 0.01 mV and 0.08 mV resulted the amount reduction of the transmitting data by 18% and 81.2%, respectively. And the simulation results show that CR can reach 2.75 when the PRD value is less than 9%.

Key words: compressed sensing (CS), dynamical model, synthetic ECG, residuals