Study on the Application of Compressive Sampling for Electrocardiogram Signals

An electrocardiograph (ECG) is a method for interpreting electrical activity in the heart using rhythm monitoring techniques for diagnosing the patient’s heart condition. This diagnostic process requires a long-term ECG signal recording and a large storage capacity. ECG signals can be compressed and then reconstructed based on a modern compression and reconstruction theory labeled as compressed sensing a.k.a. compressive sampling (CS), which reduces storage needs. Distortion in the reconstructed ECG can affect signal waves and the computed peak to peak interval. Since heart rate variability (HRV) is sensitive to changes in ECG peak to peak intervals, the accuracy of HRV computed from the reconstructed ECG signal can be used to indicate the compression and reconstruction quality. We examine the application of CS for ECG signals. The research methods include data preprocessing, ECG signals transformation using discrete Fourier and cosine transforms, compression and reconstruction using orthogonal matching pursuit and regularized multiple measurement vectors focal underdetermined system solver (RM-FOCUSS) algorithms, HRV analysis in the time domain, and evaluating RM-FOCUSS performance using normalized root sum squared error (NRSSE) for ECG signals and normalized root squared error (NRSE) for each parameter in HRV analysis. The numerical results show good NRSSE for ECG signals, ranging from 0.024215 to 0.083305, and good NRSE for HRV analysis, ranging from 0.030856 to 0.339584, for compression rates of 50% to 20% and three different measurement matrices. This indicates that the CS theory application is effective in compressing ECG signals.