HRV-Based Detection of Fear of Heights in a VR Environment

In the present study, an electrocardiographic (ECG) -based system is proposed for the classification of three levels of fear of heights. A virtual reality (VR) environment was employed for the gradual exposure of the participants to the fear arousing stimuli. The VR scenario consists of a canyon in which a wooden lift brings the subjects to three different height levels. 20 subjects participated in the experimental activities and carried out three experimental sessions. The use of psychometric tools like the Acrophobia Questionnaire (AQ) and the Subjective Unit Of Distress (SUD) allowed to carry out an initial screening of the sample to assess the severity of fear of heights and the effectiveness of the VR environment in the induction of fear. According to the AQ and SUD, three clusters of subjects with different levels of acrophobia severity were identified. A 1-lead ECG recording was acquired during the exposure to the eliciting VR environment. Hearth rate variability (HRV) -related features were extracted from the ECG signal, specifically linear (statistical and geometric) and nonlinear features. Those features were input to different classifiers for discriminating three-levels of fear. Domain adaptation methods resulted effective in improving the generalizability of the results. On the contrary, clustering the subjects according to their acrophobia severity level did not impact on the classification performances. Average accuracies of 40.9±5.9 in the inter-subject setting and of 43.8±7.6 in the intra-subjects setting were achieved by employing the Linear Correlation Alignment (CORAL) DA method.