Remote Heart Rate Estimation by Pulse Signal Reconstruction Based on Structural Sparse Representation

In recent years, the physiological measurement based on remote photoplethysmography has attracted wide attention, especially since epidemic of COVID-19. Many researchers paid great efforts to improve robustness illumination and motion variation. Most existing methods divided ROIs into many sub-regions extracted heart rate separately, while ignoring fact that rates from different are consistent. To address this problem, in work, we propose a structural sparse representation method reconstruct pulse signals (SSR2RPS) estimate rate. The (SSR) considers chrominance should have similar combined dictionary. Specifically, firstly eliminate signal deviation trend using adaptive iteratively re-weighted penalized least squares (Airpls) for each sub-region. Then, conduct dictionary, which is constructed considering pulsatility periodicity Finally, obtain reconstructed with power spectrum analysis. experimental results public UBFC COHFACE datasets demonstrate significant improvement accuracy estimation under realistic conditions.