Detection of Break in Presence in Full Body Illusion using Machine Learning

Break in presence (BIP) is a concept that refers to that condition in which the person immersed in a virtual reality (VR) environment disengages from it, due to various factors, and focuses more on the real environment. BIP occurs due to certain factors that may be related to errors in the sensory data of the virtual reality environment (VRE). These errors can influence the multisensory integration process by generating a multisensory conflict, which results in BIP phenomena. Several studies have identified a correlation between BIP and the physiological (e.g., electrocardiogram/ ECG) responses of people, showing that there is a correlation between these two aspects. In this study, we used the physiological (ECG) data collected in a previous study, in which participants were involved in a Full Body illusion (FBI) experiment, i.e. a paradigm that uses VR to generate a synchronous multisensory stimulation condition and an asynchronous multisensory stimulation condition. The FBI asynchronous stimulation condition can generate a multisensory conflict that is the same phenomenon underlying BIP. In this study, we used these physiological (ECG) data, collected throughout the FBI, to train different machine learning (ML) models to be able to detect the multisensory conflict underlying BIP from ECG tracings alone. The results showed that two ML models, Decision Tree and Random Forest, are able to detect the multisensory conflicts underlying BIP.