Analysis of electrocardiogram (ECG) signals for human emotional stress classification

Abstract

Electrocardiogram (ECG) signal significantly reflects autonomic nervous system (ANS) activities during emotional stress changes. Undeniably, a variety of valuable information can be extracted from a single record of ECG signal. Audio-visual stimuli are selected arbitrarily for the laboratory experiment in order to induce emotional stress on 5 healthy subjects. Time domain features: heart rate (HR), mean R peak amplitude (MRAmp), and mean R-R intervals (MRRI) are extracted from ECG signals and mapped into emotional stress classification using K-nearest neighbor (KNN) and Support Vector Machine (SVM). Classification performance has been investigated on three different sets of training and testing feature vector. The best mean classification accuracy for HR, MRAmp and MRRI in three classes is 66.49%, 56.95% and 61.52%, respectively and two classes are 77.69%, 61.48% and 60.21%, respectively. These results indicate that, the proposed methodology have a higher significance on distinguishing the emotional stress state of the subjects. © 2012 Springer-Verlag.