Monitoring heart rate (HR) during exercise can be used not only for early detection of a change in physical condition but also as an indication of the exercise training condition. Photoplethysmography (PPG) is a simple and non-invasive method for HR sensing, but its use during exercise is difficult, since the PPG output is contaminated with motion artifact (MA). Especially for vigorous exercise, MA has the same frequency component as that of blood volume pulses (BVPs) in the PPG output, and no linear MA cancellation technique such as band pass filtering (BPF) can work effectively. This paper proposes a PPG-based MA canceling HR sensor. It is equipped with two sensors ; one is a normal PPG sensor with a light emitting diode (LED)/photo detector (PD) that contacts the skin to detect BVPs andMA, andthe other is an MA sensor with a LED/PD that does not contact the skin to detect only MA. Applying the outputs from the two sensors to an adaptive filter, the BVPs can be obtained easily. We evaluated the performance of the proposed HR sensor in ten subjects who performeda series of resting, walking, running andjumping. The results revealedthat the proposedHR sensor, together with an outlier rejection method, achieved root mean square errors (RMSEs) of 7.1 bpm, 6.4 bpm and6.1 bpm for walking, running andjumping, respectively, when using a Holter monitor as the reference. These results imply a reduction of RMSE by a maximum of 83% compared to no MA cancellation.
