Modeling the Intent to Interact With VR Using Physiological Features

Area of Science:

• Human-Computer Interaction
• Neuroscience
• Virtual Reality

Background:

• Mixed-Reality (XR) technologies aim for natural user experiences (UX) comparable to real-world interactions.
• Achieving zero-lag and minimal mental load in XR is challenging due to technical constraints like motion-to-photon latency and inaccurate gesture recognition.

Purpose of the Study:

• To explore the use of physiological signals to model user intent for interaction within virtual reality (VR) environments.
• To overcome limitations of interactive devices lagging behind user intention through accurate prediction of interaction intent.

Main Methods:

• Computed time-domain features from electroencephalography (EEG) and electromyography (EMG) recordings during a VR grab-and-drop task.
• Cross-validated a Linear Discriminant Analysis (LDA) model using EEG, EMG, and combined EEG-EMG features.

Main Results:

• Classifiers detected pre-movement states indicating user intent to interact with virtual objects above chance level (EEG: 62% ± 10%, EMG: 72% ± 9%, EEG-EMG: 69% ± 10%).
• Leveraged features offer low computational cost and enable fast decoding of user mental states.

Conclusions:

• This work advances the classification of user interaction intent, crucial for high temporal resolution and rapid detection.
• Facilitates natural XR experiences by enabling zero-lag adaptive interfaces that respond instantly to user intentions.