Physiological sensing systems and AI-based signal analysis in immersive virtual reality learning and training: A scoping review.

Abstract

This scoping review examined research on physiological sensing systems and AI-based signal analysis in immersive virtual reality (IVR) learning. It compared studies in terms of physiological modalities, HMD integration, calibration and synchronization, artifact management, feature engineering and representation learning, fusion strategies, validation design, and deployment readiness. We searched for studies published between 2016 and 2026 and included 19 empirical studies. In the 19 included studies, eye-tracking data (ETD, 13/19) and central nervous system signals (CNS, 11/19) were the most frequently reported modalities. CNS + ETD was the most frequently reported dual-modality configuration. Peripheral physiological signals (PPS, 3/19) were less frequently represented, and studies using electrodermal activity (EDA) were especially limited within the included corpus. The included studies covered the main stages of acquisition, preprocessing, feature engineering and representation learning, fusion, and analysis and modeling. Their targets included cognitive states, affective states, performance, attributes, and response patterns. Within this corpus, early fusion and traditional machine learning were frequently reported approaches. Reporting on sample-level processing, label construction, and train-test boundaries was incomplete in several studies. As a result, the interpretation and generalizability of reported findings depend on the specific study context. Overall, the included studies collectively outline an initial technical pathway, but evidence for reproducible, verifiable, and deployment-oriented real-time closed-loop IVR physiological sensing systems remained limited.