Machine learning approach and physiological parameters evaluation: a case study on data acquired by Cinematic Virtual Reality scenarios

In response to the global concern over road traffic accidents, particularly among young drivers, there is increasing interest in innovative educational interventions to promote safer driving behaviors. Traditional media campaigns often fail to engage audiences meaningfully. This study, part of a broader research initiative, investigates the effects of Cinematic Virtual Reality (CVR) on participants' physiological responses, specifically focusing on heart rate variability, to assess the effectiveness of fear-based and positive reinforcement scenarios influencing road safety behaviors. A custom-developed CVR application delivered immersive 360° driving scenarios, one based on fear and the other on positive reinforcement, via a Meta Quest 2 headset. A total of 95 participants, aged 18 to 24 and all holding valid driver's licenses, were randomly assigned to one of the two experimental conditions. The study assessed the impact of these approaches on participants' physiological responses by recording electrocardiogram (ECG) data using the BioSignalPlux system. HRV parameters, indicators of stress and arousal, were analyzed to distinguish the physiological effects of the two scenarios. Time-domain, frequency-domain, and nonlinear HRV features were extracted and used for classification. Machine learning algorithms were employed to evaluate the system's performance in differentiating between the two experimental groups. The classification accuracy, sensitivity, and specificity were assessed using 10-fold cross-validation. Results showed that Linear Discriminant Analysis (LDA) achieved the highest classification accuracy.