Abstract:

Abstract The objective of the present study is to develop an expert diagnostic system for hypertension detection using a photoplethysmogram (PPG) Signal, which overcomes the limitation of the existing model in which accuracy is dependent on quality PPG signals, acquiring high quality can pose a challenge in real-life situations. Our proposed expert diagnostic system uses multi-domain features obtained by combining morphological features and features extracted from the decomposed PPG signal using Variational Mode Decomposition (VMD). ReliefF feature selection is used to select the top 16 features from each feature extraction approach. It is found from the comparative analysis that an expert diagnostic system based on multi-domain features showed significant improvement over the model based on single-domain features. and found to be more immune to noisy PPG signals compared to the single domain-based classification model. A variety of classifiers used are Gradient Boosting Classifier and multilayer perceptron The highest classification accuracy of F1 score for the category normal vs prehypertension, normal vs hypertension type 1 and normal vs hypertension type 2 is found to be 100%, 100% and 100% respectively using hybrid feature and  MLP.

Keywords: