Assessing the accuracy of photoplethysmography for wearable heart rate monitoring based on body location and body motion in uncontrolled outdoor environments

Abstract

Introduction

Reflective photoplethysmography (PPG) is the dominant method for heart rate (HR) monitoring consumer wearables. However, motion artifacts and sensor placement impact the accuracy of these HR measurements. Here, we present a study on how these two factors affect the accuracy of PPG-based HR measurements and compare them to ground-truth measurements from electrocardiography (ECG). Our study investigated these measurements in outdoor environments outside controlled laboratory settings.

Methods

Our study collected a dataset of 16 participants, each for 13 hours wearing four reflective PPG sensing devices placed at the forehead, sternum, ankle (supramalleolar), and wrist. Participants traveled by train from downtown Zurich to the Jungfraujoch railway station at 3,460 m above sea level in the mountains. PPG measurements were obtained using a MAX86141 optical analog front-end coupled with a green LED and photodiode from an SFH7072 module. Motion was quantified using two accelerometers (LIS2DH, ADXL355). A Lead I ECG was obtained by the device at the sternum using a MAX30003 biopotential sensor connected to gel electrodes on the chest. All devices were synchronized by aligning recorded signals post-hoc (33ms accuracy, Meier & Holz, 2023). The HR was extracted from the ECG by time-domain peak detection. The HR extraction from PPG was both performed by time-domain peak detection and frequency-domain analysis. The HR was computed every 5 seconds (30 seconds window size) resulting in 152,000 HR measurements across the whole dataset.

Results

The forehead and chest locations exhibited the highest HR measurement accuracy (median error 7.1% and 7.7%, respectively), while lower accuracies were observed for ankle and wrist placements (9.9% and 18.4% error). At rest, all median errors were below 5% while movements influenced readings at all locations negatively. Adjusted for motion, the HR obtained from the forehead sensor was most accurate. In terms of processing method, time-domain analysis produced better accuracy during periods of low motion while frequency-domain analysis was more reliable during movements.

Discussion/Conclusion

The accuracy of PPG-based HR measurements in uncontrolled outdoor settings is affected both by body location and motion artifacts with a clear ranking of site suitability: forehead >> chest >> ankle >> wrist. This is consistent with prior studies in controlled environments, though our study found a higher impact of motion than body location on HR accuracy (Longmore et al., 2019). This may be because participants’ motions in uncontrolled environments are more irregular and diverse, resulting in deteriorated signal quality. Our study shows the importance of further investigations in everyday conditions on the path toward more reliable PPG-based HR monitoring in wearable devices.

References

Longmore, S. K., Lui, G. Y., Naik, G., Breen, P. P., Jalaludin, B., & Gargiulo, G. D. (2019). A comparison of reflective photoplethysmography for detection of heart rate, blood oxygen saturation, and respiration rate at various anatomical locations. Sensors, 19(8), Article 1874. https://doi.org/10.3390/s19081874

Meier, M., & Holz, C. (2023). BMAR: Barometric and Motion-based Alignment and Refinement for offline signal synchronization across devices. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 7(2), Article 69. https://doi.org/10.1145/3596268