During the execution of the research project, the following scientific and technical tasks will be accomplished:
- optimization of deep artificial neural networks for a cuffless blood pressure measurement system in order to improve computational efficiency and noise robustness by simplifying and retraining the fully connected layer of the feedforward neural network;
- development of a convolutional neural network–based method for extracting photoplethysmographic (PPG) signals from a human face through video image processing;
- development of preprocessing methods for suppressing noise and artifacts in the extracted photoplethysmographic signals from images;
- performing a comparative analysis of blood pressure estimation using the developed methods, including approaches based on photoplethysmographic signal processing, bispectral analysis, Fourier and wavelet spectra, discrete cosine transform (DCT) of photoplethysmographic signals, video-based photoplethysmography, as well as convolutional neural network–based methods.