Prediction of daily mental stress levels using a wearable photoplethysmography sensor

Abstract

In this study, we investigated a feasibility to predict a daily mental stress level from heart rate variability (HRV) using a photoplethysmography (PPG) sensor in the wristband-type wearable device. We performed an experiment in which each participant measured their PPG signals for 30 s using the wristband three times a day for a week. The recorded signals were transmitted to and stored at a smartphone via the Bluetooth link by custom-made software. At the end of each day, participants also self-evaluated their mental stress level using the perceived stress scale (PSS). A preprocessing procedure was used to remove environmental artifacts in the PPG signal and HRV was estimated from the PPG signal by the detection of PPG peaks. We then extracted a low-frequency (0.04Hz-0.15Hz) / high-frequency (0.15Hz-0.4Hz) feature of HRV using the autoregressive (AR) model. A linear regression model predicted the self-reported mental stress level from the HRV features. Prediction accuracy was 86.35% on average across the participants. The proposed method could demonstrate a feasibility of developing a mobile health solution that predicts a personal mental stress level using HRV measured by a wristband PPG sensor.