Real-Time Flow State Analysis in Game: A Non-Intrusive System Based on Eye Feature Detection

Flow is a key indicator for deep immersive state. Current methods for detecting flow often rely on intrusive physiological devices (e.g EEG, HRV), requiring specialized equipment and complex data processing, with no capability for real-time flow state visualization. These limitations hinder the possibility to analyze the relationship between game settings and flow state fluctuations. To address this, we utilized an eye-tracking device (Tobii glasses) to record eye movement features in games under different flow states, creating a dataset to identify significant eye movement patterns correlated with flow state. Based on this, we developed a non-intrusive eye-tracking system that captures and analyzes specific eye features through a custom-designed algorithm, providing real-time feedback on flow states. Our system achieved a flow state detection accuracy of 85.3%, with a 412 ms detection delay, outperforming existing methods. This innovation offers a powerful tool for exploring the relationship between flow and game design, and further unlocking the biophysical mechanisms of flow.