A smartphone-based finger-pressing blood pressure monitoring system using a fusion model

Existing cuffless blood pressure (BP) measurement methods based on photoplethysmography (PPG) typically require frequent cuff-based calibration, limiting their adoption in daily life. Recently, finger-pressing methods have become a promising approach to provide calibration-free BP monitoring. This study presents a smartphone-based finger-pressing BP monitoring system that includes a user-friendly system integration for simultaneous acquisition of three multimodal signals, and an Android application that provides real-time visual guidance for pressure application. A novel fusion model is proposed, combining the oscillometric envelope method (OMWE) and pulse arrival time (PAT) estimation through a joint loss function. We further adopt an optimized Adam gradient descent (AGD) algorithm for the parameter estimation in the embedded device. By tested on 20 healthy volunteers, the proposed fusion model showed improved accuracy compared to the traditional OMWE model. Combined with the nonlinear least squares fitting algorithm (NLS), the fusion model yielded mean absolute errors (MAEs) of 8.4 mmHg for systolic BP (SBP) and 15.0 mmHg for diastolic BP (DBP). When implemented in the embedded device using optimized AGD algorithm, the fusion model maintained its favorable accuracy, with MAEs of 9.6 and 16.1 mmHg for SBP and DBP respectively. The study also investigates operational factors attributing to compromised DBP estimation performance, and highlights the need for improved interaction design.