Research map / Label-efficient and robust learning / Recognition and long-tail
Gaze-Based Interaction Adaptation for People with Involuntary Head Movements (Student Abstract)
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes HMDSA, an adaptation algorithm combining head-pose-based gaze correction and speed-controlled head position smoothing, which a two-participant study found shortened icon selection time and raised accuracy for users with involuntary head movements.
Problem
Eye-tracking and gaze-based interaction are unreliable for people with dyskinetic cerebral palsy, severe intellectual disabilities, or other involuntary head movements, because the eye tracker assumes a stable head.
Approach
A head pose monitor adjusts the detected gaze point using 3D head position and y-axis rotation, and a gaze speed controller smooths head position updates via a speed-dependent impact factor, stabilizing gaze points on screen without learned models.
Key results
- Selection time decreased from 7.72 s (control) to 4.79 s (experimental) with adaptation (ANOVA p < .001, F(1,398)=82.39, eta2=.17).
- Selection accuracy increased from 51% (control) to 92% (experimental) (p < .001, F(1,398)=103.41, eta2=.21).
- Study used two participants with involuntary head movements, 100 selection tasks per trial, and N=400 observations in the statistics.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add head movement adaptation | Gaze-based interaction with HMDSA · 1-out-of-8 icon selection | selection time (s): 7.72 (4.32) -> 4.79 (1.52) | Helped |
| Add head movement adaptation | Gaze-based interaction with HMDSA · 1-out-of-8 icon selection | selection accuracy (%): 51 (51.1) -> 92 (27.2) | Helped |
Limitations
- Lower performance when users wear glasses.
- Head pose monitor performs better for horizontal than vertical head movement.
- Very small sample size (two participants) limits generalisation.
Experimental setup
Metrics: selection time (s) selection accuracy (%)
Baselines: Control condition without adaptation
Closest papers by meaning
- EyEar: Learning Audio Synchronized Human Gaze Trajectory Based on Physics-Informed Dynamics (AAAI 2025 · similarity 0.82)
- Gaze from Origin: Learning for Generalized Gaze Estimation by Embedding the Gaze Frontalization Process (AAAI 2024 · similarity 0.81)
- CLIP-Gaze: Towards General Gaze Estimation via Visual-Linguistic Model (AAAI 2024 · similarity 0.81)
- Learning from Observer Gaze: Zero-Shot Attention Prediction Oriented by Human-Object Interaction Recognition (CVPR 2024 · similarity 0.81)
- GazeFlow: From Human Gaze Behavior to Generative Egocentric Gaze Prediction (NeurIPS 2026 · similarity 0.80)
- ST-DiffEye: Diffusion-based Continuous Gaze Generation via Joint Scanpath-Trajectory Modeling (NeurIPS 2026 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.