Towards Stereoscopic Vision: 3d Gaze Estimation from Eeg Signal

Abstract

Gaze estimation typically relies on a complex interplay of cameras, lighting conditions, and detailed eyeball models to detect eye movements. These approaches, however, face challenges with depth estimation, limiting its usage in 3D environments like virtual reality (VR). In this study, we explore an alternative method using the brain-computer interface (BCI), which allows us to decode human intentions directly from brainwave data, providing a more intuitive approach to gaze estimation. We first conducted a preliminary survey to gauge public views of eye-tracking and BCI technologies, revealing positive attitudes towards BCI's potential for gaze detection. Building on this insight, we collected an EEG dataset in a VR environment, where participants were exposed to various 3D stimuli. We trained a model to predict the continuous distribution of spatial attention in 3D space. This approach allows attention intensity calculation at any point and enables precise point of regard (PoR) estimation through backpropagation. To our knowledge, this work is the first to explicitly map brain activity to visual field objects within the context of BCI-driven visual cognition.

Keywords: Gaze Estimation, 3D, Brain-computer interface, Spatial Attention Prediction, Virtual Reality (VR)