Per Subject Complexity in Eye Movement Prediction

2024-12-31

AI Summary

To address the modeling challenge in virtual reality eye-movement prediction arising from inter-subject variability, this paper introduces the novel concept of “subject-wise complexity,” establishing an interpretable, event-to-subject evaluation metric that systematically attributes sources of individual differences and explores mitigation strategies. Methodologically, we integrate a lightweight LSTM, a temporal Transformer (TST), and an oculomotor physiological model encapsulated via Kalman filtering (OPKF), adopting a sample-to-event evaluation paradigm. All three models consistently rank subjects by difficulty, validating the universality of the proposed complexity metric. Correlation analysis further identifies key determinants at both physiological (e.g., saccadic latency) and behavioral (e.g., fixation stability) levels. This work provides a theoretical foundation and a transferable evaluation framework for personalized eye-movement modeling.

Technology Category

Cognitive Modeling & Cognitive Systems: Simulating Human BehaviorComputer Vision: Motion & TrackingMachine Learning: Calibration & Uncertainty Quantification

Application Category

User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating successSecurity and Privacy: Large-scale security measurementsSearch and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Eye movement prediction is a promising area of research to compensate for the latency introduced by eye-tracking systems in virtual reality devices. In this study, we comprehensively analyze the complexity of the eye movement prediction task associated with subjects. We use three fundamentally different models within the analysis: the lightweight Long Short-Term Memory network (LSTM), the transformer-based network for multivariate time series representation learning (TST), and the Oculomotor Plant Mathematical Model wrapped in the Kalman Filter framework (OPKF). Each solution is assessed following a sample-to-event evaluation strategy and employing the new event-to-subject metrics. Our results show that the different models maintained similar prediction performance trends pertaining to subjects. We refer to these outcomes as per-subject complexity since some subjects' data pose a more significant challenge for models. Along with the detailed correlation analysis, this report investigates the source of the per-subject complexity and discusses potential solutions to overcome it.

Problem

• Eye Movement Prediction
• Individual Variability
• Virtual Reality Headset Optimization

Innovation

• Eye Movement Prediction
• Individual Variability
• LSTM vs TST vs OPKF