Blood pressure estimation from phonocardiographic signals recorded with a smartphone and an electronic stethoscope

Phonocardiography (PCG) is a non-invasive diagnostic method that enables the analysis of heart acoustic activity. This study investigates the feasibility of estimating blood pressure based on PCG signal parameters recorded using a smartphone microphone and an electronic stethoscope. A database of 220 ten-second heart sound fragments was created, each linked to simultaneous blood pressure measurements. Noise detection and signal segmentation were performed using a Hierarchical Segmental Hidden Markov Model (HSMM). Selected acoustic features were used to build predictive models employing machine learning algorithms, with Random Forest achieving the highest correlation with actual pressure values (r > 0.8). The results suggest that smartphone recordings can be as effective as those obtained with professional equipment, paving the way for mobile diagnostic tools.