Profiling the propagation of error from PPG to HRV features in consumer wearable devices

In this white paper we explore the reliability of the heart rate sensor in the Microsoft Band v2, a wrist worn consumer wearable device, and the propagation of error from heart rate to heart rate variability features. We show that motion artifacts account for most of the error, and that is possible to filter unreliable portions of heart rate data, using the accelerometer sensor present in the wearable device.

We show that the noise has strong frequency components in the HF band, making the SVI HRV feature unreliable.

We also show that long windows are not needed to extract accurate estimations of time domain HRV features.

About the author
Davide Morelli, PhD

Davide leads BioBeats’ engineering team. He is a specialist in the intersection between Artificial Intelligence and music, and previously ran a successful distributed software consultancy (Parser) for ten years. He is also an active composer with a PhD in Computer Science that focuses on energy models for algorithms that illuminate correlations.