A novel feature extraction algorithm for on the sensor node processing of compressive sampled photoplethysmography signals

This paper proposes an efficient method for extraction of heart rate (HR) and heart rate variability (HRV) information from compressively sampled photoplethysmogram (PPG) signals. The proposed approach utilizes Lomb-Scargle periodogram to perform Least-squares spectral analysis to extract the spectral content from compressively sampled PPG signals. The spectrum thus obtained is used to estimate the average HR and HRV. Simulation results demonstrate that the average HR estimated using the proposed method is accurate within ±5 beats per minute (bpm) while HRV exhibits a correlation coefficient of > 0.90 at 30x compression ratio (CR) compared to time domain HR and HRV estimation performed in Nyquist sampled PPG signals. This facilitates embedded ultra-low power, on-the sensor node feature extraction from compressively sampled PPG signal without requiring complex reconstruction techniques.