BPM Analysis Under Motion

Abstract

Measuring heart rate using photoplethysmographic (PPG) signals under an intensive motion is a tough task. The PPG signals will be greatly affected by the subject’s movements. This project aims at develop an algorithm to estimate a person’s heart rate during exercise. Few sets of 300 seconds PPG signals, which were recorded for a person running at different speeds, were studied. In the algorithm, the PPG signals were divided into many overlapping time frames. For each time frame, the PPG signals were transformed to frequency domain using FFT. The strong frequency components undergoing harmonic analysis are used as candidates for the heart rate. Beat pulse per minute (BPM) tracking is an essential step in this algorithm, since there could be some strong motion artifacts, the tracking helps eliminate the BPMs below a threshold. In the tracking algorithm, the heart rate trend is computed to define a range for selecting a suitable candidate for computing the BPM. The trend helps eliminate candidates outside the range and extrapolate a value when there is no candidate within the range. The algorithm was evaluated for different data sets. Experimental results show that the algorithm can provide good heart rate estimation under motion.