Human Blood Pressure Measurement Using Machine Learning Strategy

A technique based on a machine learning approach is suggested and studied for blood pressure measurements. The proposed technique uses a noninvasive cuffless approach for blood pressure evaluation. In order to extract blood pressure data using this noninvasive cuffless method, pulse wave velocity or pulse wave travel time (PTT) are estimated by both signal processing of electrocardiogram (ECG) and photoplethysmogram (PPG) data records. For study performed by computer simulations, the ECG and PPG records were taken from an open database. Errors arising both for systolic and diastolic arterial pressure evaluation were estimated. Computer simulation results indicate that using machine learning strategy and using only PTT parameters provide a considerable decrease in root mean square errors both for systolic and diastolic human blood pressure data.