Human activity recognition using a single optical heart rate monitoring wristband equipped with triaxial accelerometer

This paper investigates activity monitoring using a single wrist-worn optical heart rate monitoring sensor that is equipped with a triaxial accelerometer. Wearing accelerometers on the wrist provides more convenience and therefore improved wear-time compliance compared to other measurement sites. Reliability of wrist acceleration for activity monitoring has been addressed in former research. However, integration of wrist acceleration with physiological signals has not been comprehensively explored yet. We investigated a variety of home-specific activities (sitting, standing, household, and stationary cycling) performed by 20 male participants. Random Forest (RF) and Support Vector Machines (SVM) were applied for activity classification. Various features calculated from acceleration, heart rate (HR), and heart rate variability (HRV) were used as classified inputs. Results of leave-one-subject-out cross-validation showed 89.2% and 85.6% average recognition accuracies for RF and SVM, respectively. HR and HRV features improved the classification rates of high-intensity cycling by 8% for RF and 7% for SVM.