Federated Adaptive Learning for Personalized Anxiety Detection in Virtual Reality Therapy: Enhancing Privacy and Accuracy

Abstract

Virtual reality (VR) therapy has emerged as a promising tool for treating anxiety disorders by immersing patients in controlled environments that simulate anxiety-inducing situations. However, detecting and personalizing anxiety levels in real-time requires access to sensitive physiological and emotional data, raising privacy concerns. This paper explores the integration of federated learning (FL) with adaptive machine learning (ML) models to develop a personalized, privacy-preserving anxiety detection system in VR therapy. By leveraging decentralized data training through FL, the system enhances user privacy while maintaining high detection accuracy and improving over time based on individual responses. We propose a novel framework, evaluate its performance, and discuss the potential for broader applications in emotion recognition and therapy optimization.