Uncertainty of heart rate variability measured through a wearable device during office activities

This paper shows the results of an experimental campaign conducted to quantify the measurement uncertainty of heart rate variability (HRV), measured through a commercial smartwatch, while participants were performing office activities. The selected activities (sitting, moving the mouse, writing a text on the laptop, handwriting, walking, walking upstairs/downstairs) are subjected to motion artifacts (MA) that can negatively impact the HRV signal. Measurement uncertainty for each activity has been computed comparing the HRV acquired by the smartwatch with a reference sensor, i.e. multi-parametric chest belt BioHarness 3.0. Activities like resting and moving the mouse exhibit an uncertainty of +/- 60 ms and +/- 90 ms, while activities such as handwriting, walking and walking upstairs/downstairs increase the measurement uncertainty from +/- 215 ms up to +/- 530 ms. These results are the preliminary step of a major study developed to measure the well-being of occupants using also HRV signals. In this context, accelerometer data, collected by the smartwatch, could possibly classify the activity performed by the workers through artificial intelligence (AI) algorithms; then, the HRV measurement uncertainty associated with the recognized activity could be used as a support for the discrimination of specific corrupted HRV segments, for minimizing the uncertainty in the HRV signal caused by MA.