A Lightweight Method for Non-Invasive Blood Pressure Estimation Using PPG and Hjorth Parameters for Wearable Devices

This work proposes a lightweight, non-invasive approach for estimating systolic and diastolic blood pressure (BP) from photoplethysmography (PPG) signals, leveraging Hjorth parameters as core features. By combining Hjorth descriptors—activity, mobility, and complexity—with conventional time-domain features, the method achieves high predictive accuracy while maintaining low computational complexity, making it suitable for integration into wearable health monitoring systems. The model was trained and evaluated on data from 85 subjects extracted from the publicly available University of California Irvine (UCI) repository, derived from the MIMIC-II database. The proposed approach achieved mean absolute errors (MAE) of 3.53 mmHg for systolic and 2.15 mmHg for diastolic BP. These results not only meet the performance requirements of the Association for the Advancement of Medical Instrumentation (AAMI), but also achieve Grade A classification under British Hypertension Society (BHS) standards. Offering performance comparable to more complex state-of-the-art models, this method stands out for its efficiency and ease of deployment on resource-constrained embedded platforms, representing a promising solution for accessible, continuous BP monitoring in real-world, low-resource healthcare environments.