Validation of Heart Rate Estimation from Photoplethysmograph and Accelerometer Recordings During Post-Exercise Relaxation

Abstract

Photoplethysmogrpah and accelerometer hardware have reached the stage at which the sensors are easily affordable and suitable for self-application as well as a part of multiparametric sensing schemes in health monitoring. However, the non-supervised extraction of biomarkers, as simple as heart rate, remains a challenge, largely due to the large variety in pulse shapes and poor signal quality when measured under stress. While a large body of work, including database generation, has been done for signals recorded distally at fingers, wrist, toes and temporal artery, the data recorded at distal arteries such as carotid, are lacking. Here we use the SensSmartTech multiparameter database to test the performance of the PPG and ACC in HR estimation at rest and during relaxation after the activity. PPG was recorded at carotid and brachial arteries and ACC in the vicinity of the precordial V4 electrode. HR is calculated using the available extrema-search functions. Results indicate excellent agreement with the reference HR extracted from the ECG signal, with the RMSE on both sensors below 3 bpm. The accuracy of the ACC measurement increased after the activity, while that of the PPG decreased, nevertheless remaining within the above error margin.