Deriving blood pressure from photoplethysmogram waveform using SVM-based machine learning algorithm

Blood pressure (BP) is a bio-parameter that can directly indicate hypertension and the health complications following it. Traditional methods of measuring blood pressure involve cuff-based measurements (manual or digital sphygmomanometers), which can only take place when the patient is at rest, and each measurement takes about 2-3minutes. Also, BP cuffs are not one-size-fits-all, making them non-inclusive. Recent advancements resulted in a number of devices for continuous measurement of bio-parameters, such as heart rate (HR), heart rate variability (HRV), respiration rate (RR), BP, etc. In this paper, the design of a continuous monitoring of Blood pressure device is proposed that solely uses photoplethysmogram (PPG) signals. The proposed design uses a support-vector regression model (SVR) model to predict blood pressure with PPG wavelets as input, that enable real-time and computationally efficient prediction of BP. Simulation results show that the proposed model predicts the for systolic BP (SBP) and diastolic BP (DBP) with mean absolute error 5.3 ±3.7 mmHg and 13.4±2.1 mm Hg respectively.