Towards an Eye-Brain-Computer Interface: Combining Gaze with the Stimulus-Preceding Negativity for Target Selections in XR

Title of the Paper

Towards an Eye-Brain-Computer Interface: Combining Gaze with the Stimulus-Preceding Negativity for Target Selections in XR

Paper Information

• Research Area: Cutting-edge research on integrating gaze tracking with Brain-Computer Interfaces (Eye-Brain-Computer Interface, EBCI) to address user interaction challenges in immersive virtual reality scenarios.
• Keywords: Spatial computing, eye tracking, gaze interaction, brain-computer interface, EEG, target selection, menu selection, assistive technology, Midas touch (false selection issue).

Research Background and Problem Statement

• Identified Problems or Challenges:

  1. While gaze interaction for target selection is intuitive, it is prone to the Midas touch problem (the system mistakenly interprets the user's gaze as a selection signal).
  2. Current solutions often rely on additional physical triggers (e.g., button presses or blinking), which increase user burden and diminish immersion.
  3. Common brain-computer interfaces (e.g., P300 event-related potentials, steady-state visual evoked potentials, motor imagery) involve high user workload, cognitive fatigue, and slow interaction speeds, limiting their applicability.

• Significance of the Problem:

  1. In immersive virtual reality (XR) and augmented reality (AR) scenarios, seamless and unconscious selection methods are critical for user experience.
  2. "Contactless" interaction methods have broad applications in assistive technology and everyday HCI.

• Research Motivation and Related Work:

  • Recently, "Passive Brain-Computer Interfaces" (Passive BCI) have introduced concepts like Stimulus-Preceding Negativity (SPN), showing potential to simplify user task burdens.
  • Existing studies on SPN are mostly based on 2D scenarios, lacking exploration in 3D immersive environments. The complexity of XR scenarios, including head movements and field-of-view changes, significantly increases research challenges.
  • Furthermore, previous studies have not clearly distinguished whether SPN is influenced by user intention or feedback information itself.

Proposed Solution

• Proposed Method or Solution:

  1. Combine eye tracking with SPN (recorded via EEG) to decode user selection intentions.
  2. Experimentally validate whether SPN results from the user's expectation of feedback, thereby avoiding the Midas touch problem.

• Innovative Contributions:

  • First exploration of SPN in 3D immersive VR environments, extending research beyond 2D constrained scenarios.
  • Introduced core causal analysis of feedback and user intention, clarifying that SPN is driven solely by selection intention.
  • Investigated the effect of target familiarity on accelerating SPN, exploring its potential in optimizing BCI time windows for everyday applications.

• Implementation Steps and Key Techniques:

  1. Design user experiments where participants interact with targets via "gaze dwell" (750ms trigger), including three conditions: intent-to-select, intent-to-observe without feedback, and intent-to-observe with feedback.
  2. Use Tobii eye-tracking integrated into a VR headset to record gaze positions in real time, while BioSemi ActiveTwo records 64-channel EEG and EOG signals.
  3. For data analysis, apply Independent Component Analysis (ICA) to remove eye movement artifacts, and use event-related potential (ERP) analysis to extract SPN signals from -750ms to 0ms.
  4. Perform one-way ANOVA and Mass Univariate Analysis to explore SPN significance under different task conditions.

Research Outcomes

• Specific Findings:

  1. The experiment validated for the first time that SPN can significantly distinguish selection intention from non-selection intention, with a notable negative potential (-4 μV) observed during user gaze at the target (intent-to-select).
  2. Confirmed that SPN is driven solely by the user's expectation of feedback selection, rather than the feedback itself.
  3. Found that target familiarity may accelerate the emergence of SPN, offering potential pathways for optimizing BCI time windows in everyday tasks.

• Advantages over Existing Solutions:

  1. Eliminates the need for additional physical input, reducing fatigue.
  2. Passive interaction is smoother compared to BCIs based on P300 and steady-state visual evoked potentials (SSVEP).
  3. Mass Univariate Analysis provides higher spatiotemporal precision, further enhancing the robustness of SPN extraction.

• Experimental or Evaluation Results:

  1. In the intent-to-select task, significant SPN negative potentials were observed at electrodes O1, PO7, and O2, reaching statistical significance.
  2. Comparing intent-to-select and intent-to-observe conditions, significant differences in time windows were detected as early as 12 milliseconds after initial target selection.

• Limitations and Future Directions:

  1. Although the experiment preliminarily showed that target familiarity accelerates SPN, the statistical significance could not be confirmed due to the filtering approach (high-pass filter application).
  2. The experimental setup relied on high-resolution laboratory equipment, and the adaptability of online real-time artifact correction or dry electrode systems requires further evaluation.
  3. The impact of completely removing feedback on SPN generation was not studied.
  4. Differentiating SPN in real-world scenarios where "selection" and "observation" conditions are mixed remains unexplored.

Conclusion and Future Outlook

This study introduces a passive brain-computer interface method by integrating cutting-edge SPN research with eye tracking, effectively addressing the Midas touch problem in immersive XR scenarios. Future directions include developing online classification models for real-world applications to enhance system robustness and exploring more complex interaction scenarios beyond selection tasks, such as object manipulation or dynamic value adjustments. This research opens new avenues for eye-brain integrated passive brain-computer interfaces, not only advancing BCIs from assistive tools to general interaction interfaces but also providing scientific evidence to address core challenges in gaze-based interaction.