Recognition of Real-Scene Stress in Examination with Heart Rate Features

This work analyzed the autonomic reactivity of real-scene stress. Electrocardiogram (ECG) data were acquired from 16 subjects during the final examinations of a semester and after a long winter holiday. Weak stress and strong stress datasets were constructed from the ECG data of the 16 subjects. Statistical analysis showed that there was significant autonomic reactivity difference between strong stress and weak stress. The datasets of weak and strong stress were classified by using the support vector machine (SVM), and obtained the accuracy of 93.75% and 87.5% for the binary classification of the strong stress and weak stress, respectively. By using the sequential backward selection algorithm of 5-fold cross validation, the mean of the RR interval series and the mean of the absolute values of the second normalized differences of the RR interval series were selected as the optimal features for distinguishing two different levels of stress.