AI-driven motion-based emotion recognition in interactive virtual reality of equine-assisted services.

Abstract

This paper explores how technology can replicate equine-assisted services without requiring physical interaction with horses. Human-horse interaction is utilised in equine-assisted therapy to improve mental wellbeing; however, access to such interventions is limited due to the high horses’ cost and maintenance. Horses possess advanced emotional intelligence, interpreting human gestures and non-verbal cues to respond adaptively, often mirroring human emotions. By modelling these capabilities in a virtual environment using Artificial Intelligence (AI) and Machine Learning techniques, users can access the replicated services while interacting with virtual horses. A system is designed to detect participants’ emotional states reflected in their motion via motion analysis and integrate these cues into an interactive virtual reality (VR) experience to foster calmness and improve communication skills. An experiment was conducted to train an emotion-recognition model for replicating horse-human interaction in VR. Results demonstrate the system’s accuracy in detecting emotional states and the findings suggest the potential of VR-based equine-assisted services to replicate the real ones to enhance accessibility and engagement with such services.