Evaluation of the quality of electrocardiograms, seismocardiograms and gyrocardiograms based on characteristics derived from symmetric projection attractor reconstruction

Abstract

This study presents the evaluation of cardiovascular signals (electrocardiograms, seismocardiograms, and gyrocardiograms) based on features derived from symmetric projection attractor reconstruction (SPAR). We assessed five classifiers: random forests, bagged trees, gradient boosting, multilayer perceptron, and support vector machine. The highest accuracy was achieved for bagged trees in ECG signals (0.6667) and the lowest overall accuracy for the multilayer perceptron in SCG signals (0.3333). The results showed that five classifiers fed features derived from SPAR can be used to assess the quality of cardiovascular signals.