Smartphone Based BP Level Monitoring System Using DNN Model

In this paper, a novel, non-invasive, cuff-less method is proposed for monitoring BP level from Photoplethysmography (PPG) signal based on Smartphone contact video instead of additional sensors or equipment. Since most people possess a smartphone, blood pressure (BP) can be measured using the smartphone’s video through PPG signal rather than the conventional procedures, which still rely on BP measurement instruments that are not accessible to everyone. During data collection, participants placed their finger on the phone camera while the camera flashlight illuminated it. A total of 65 individuals provided fingertip videos for this study. From each subject’s fingertip videos, PPG signal was generated at a frame rate of 60 fps (frame per second). After preprocessing to remove noise and baseline drift, 40 distinct features were extracted from the PPG signal using single period, first and second-order derivatives, and Fourier analysis. The extracted features are then used as input in a DNN model to train the model to evaluate the measurement. The PPG signal benefits from 60 fps frame rate as it enables precise tracking of subtle changes in blood flow and heart rate, whereas using the arPLS algorithm for baseline correction results in a clear PPG signal by eliminating baseline drift, allowing for a more dependable analysis of blood pressure levels. This approach provides an estimated R2=0.963 and R2=0.952 accuracy in the case of systolic and diastolic BP level measurements, respectfully.