Publication

Abstract : Accurate blood pressure (BP) monitoring is crucial for cardiovascular health, but traditional cuff-based methods are often uncomfortable and impractical for continuous use. This study presents a non-invasive BP estimation approach using facial video analysis, photoplethysmography (PPG), and machine learning techniques. Regions of interest (ROIs) are extracted from facial videos to obtain PPG signals from the red, green, and blue color channels. These signals are processed using Fast Fourier Transform (FFT), Independent Component Analysis (ICA), and bandpass filtering to enhance signal quality. Power Spectral Density (PSD) analysis is then applied, and Pulse Transit Time (PTT) is computed as a key feature for BP estimation. Machine learning models, including Linear Regression, Polynomial Regression, Random Forest, XGBoost, Support Vector Regression, and Multi-Layer Perceptron (MLP), are trained to estimate systolic and diastolic blood pressure. The correlation between PTT and BP is analyzed to select the most relevant features for prediction. The proposed method demonstrates promising accuracy, providing a real-time, continuous, and non-invasive alternative for BP monitoring. This approach has the potential to improve cardiovascular health tracking, making BP measurement more accessible and convenient without the need for traditional cuff-based devices.