Comparative analysis of the quality of ECG signal reconstruction by compressed sensing at different heart rates

Objective: Compressed Sensing (CS) has emerged as an effective strategy for low-power remote monitoring of electrocardiogram (ECG) signals. However, reconstruction performance may degrade under variations in signal sparsity induced by changes in heart rate. This study systematically evaluates the influence of heart rate on CS-based ECG reconstruction. Methods: Synthetic ECG signals were generated at 30, 80, 120, 180 and 240 bpm. Reconstruction quality was assessed using Percentage Root-Mean-Square Difference (PRD) and Signal-to-Noise Ratio (SNR). Results: Increasing heart rate led to reduced sparsity and a consistent deterioration in reconstruction quality across algorithms. A three-way ANOVA indicated statistically significant effects of algorithm, compression ratio (CR), and heart rate on PRD and SNR ( p < 0 . 001 for all factors), with a significant Algorithm × CR interaction ( p < 0 . 001 ). Post-hoc Tukey analysis confirmed that the deep learning-based CSNET achieved significantly superior reconstruction compared to OMP and BP. Conclusion: Acceptable performance was maintained for CRs up to 60%, although the sensitivity to heart rate varies across algorithms. Significance: These findings provide statistically robust evidence of how heart rate modulates CS reconstruction behavior and offer practical insights for optimizing CS-based remote ECG monitoring in wearable and telehealth applications.