Abstract
Heart rate sensing is applicable for effective physical training. When sensing heart rate by means of photoplethysmography (PPG), however, motion artifact (MA) during exercise becomes larger than blood volume pulse (BVP) in the PPG output, and furthermore, the frequency component of the MA overlaps that of the BVP. As a result, conventional linear filtering techniques cannot reject the MA effectively. We propose a PPG-based heart rate sensor with MA cancellation which utilizes two kinds of sensors; one LED/PD sensor contacts the skin to detect the MA and BVP, and the other LED/PD sensor does not contact the skin to detect only the MA. Applying an adaptive algorithm to the two sensor outputs, the heart rate sensor can cancel the MA even during exercises. A session of the experiment was composed of repetition of standing still, fast-walking and running, and we applied a recursive least squares (RLS)-based MA cancellation algorithm for the two outputs. The results showed that, as compared to the reference value obtained from the Holter ECG, the root mean square error (RMSE) of the proposed sensor was 26 bpm in average, whereas the RMSE of a band pass filtering was 45 bpm in average.
