Abstract:

The objective evaluation of the degree of fear plays an important role in mental health and professional ability assessment. Physiological and exercise data in the immersive training process can improve the interface design and increase the safety of virtual training. This article used the HTC Vive head-mounted VR system to realize the immersive VR experience. OpenVR open source software package was used to collect the position information of the controller and the head, which was combined the ECG signals and body acceleration collected by the Equivital belt EQ02 Lifemonitor to realize feature extraction of various signals. In this paper, the feature ranking method and support vector regression utilizing recursive feature elimination were used to select 10 features from 50 physiological and motion features. Finally, the multivariate polynomial regression based on self-tested fear values and features achieved a binary classification with an accuracy rate of 90%, which completed the classification of fear or non-fear of the subject.

Key words: fear measuring model, virtual reality, physiological signals, movement and performance, SVR, recursive feature elimination