Recognizing cardiovascular risk from photoplethysmogram signals using ELM

Abstract

In this paper, photoplethysmogram (PPG) signals, 30 healthy and 30 pathological, are classified as `healthy' or `at risk' of cardiovascular diseases (CVDs) using extreme learning machine (ELM), a supervised learning algorithm. Additionally, two other supervised learning algorithms, backpropagation and support vector machine are used for classification to compare their results with that of ELM and hence validate its performance. Based on the results obtained, ELM gives the best accuracy, a sensitivity of 89.33% and a specificity of 90.33%, with minimum training time and minimum number of features as input.