Self-supervised Learning with Demographic Information for Cuffless Blood Pressure Estimation

Photoplethysmography (PPG) can be conveniently and continuously measured through wearable devices. Cuffless blood pressure (BP) estimation is an important application of PPG technology. Due to the complex physiological characteristics of PPG signals and the impact of individual differences on PPG measurement, incorporating demographic information, including age, gender, height, and weight, into PPG-based BP estimation can potentially achieve more accurate results. However, previous self-supervised learning methods for BP estimation ignored the impact of demographic information. Therefore, this study proposed a new self-supervised learning method that combined demographic information for BP estimation. The BP estimation accuracies were assessed with the public PulseDB dataset. The results showed that when only using PPG for BP estimation, the mean absolute errors (MAE) of the proposed method on systolic blood pressure (SBP) and diastolic blood pressure (DBP) were 5.41 and 2.27 mmHg, respectively, which outperformed those of recent methods.