PPG SİNYALLERİNİN TQWT TABANLI AYRIŞTIRILMASI YOLUYLA KAN BASINCI VE KALP ATIŞ HIZI TAHMİNİ

Abstract

Photoplethysmography (PPG) signals are getting more popular and promising for medical applications because of the non-invasive, fast, and simple recording techniques. Using PPG signals for monitoring the blood pressure (BP) and heart rate (HR) levels instead of traditional invasive and cuff-based measurement techniques is possible and continuous tracing of BP and HR levels can be accomplished with high measurement accuracies. These developments are very important and helpful, especially for people suffering from high tension and cardiac problems. In this study, we propose to use Tunable Q-factor Wavelet Transform (TQWT) for decomposing the PPG signals into sub-signals and extracting some statistical features from each of the sub-signals and main signal. Artificial Neural Networks (ANN), Random Forests (RF), and Support Vector Machines (SVM) algorithms are employed to estimate diastolic blood pressure (DBP), systolic blood pressure (SBP), and heart rate (HR) values. PPG signals, DBP, SBP, and HR values which were measured with traditional methods were obtained from the open dataset of Guilin People’s Hospital of China. This dataset includes information of 219 individuals. Each machine learning method was applied to the features separately, and the results of the regression analysis were interpreted by using the error rates and correlations between the actual and estimated values. Results show that the RF algorithm is more successful than ANN and SVM for the estimation of DBP, SBP, and HR levels.