Towards automation of dynamic-gaze video analysis taking functional upper-limb tasks as a case study
Musa Alyaman1, Mohammad Sobuh2, Alaa Abu Zaid1
1Mechatronics Engineering Department, School of Engineering, The University of Jordan, Amman, 11942, Jordan.
Computer Methods and Programs in Biomedicine
|March 23, 2021
Summary
Automating gaze data coding for upper-limb tasks significantly speeds up analysis and improves reliability compared to manual methods. This new algorithm offers a more efficient and objective approach to quantifying attention in motor control studies.
Area of Science:
- Motor control research
- Human-computer interaction
- Biomedical engineering
Background:
- Gaze behavior quantifies attention during actions.
- Manual coding of gaze data is time-consuming and subjective.
- Need for automated analysis of gaze data in motor control.
Purpose of the Study:
- To assess the feasibility of automating gaze data coding.
- To develop an algorithm for objective analysis of gaze patterns.
- Case study focused on functional upper-limb tasks.
Main Methods:
- Developed a three-stage algorithm: data preparation, processing, and output generation.
- Processed gaze data (crosshair and video) into a 25 Hz frame sequence.
- Utilized image processing and a fuzzy logic controller to identify gaze fixations in Areas of Interest (AOIs).
Main Results:
- The automated algorithm demonstrated high agreement with manual coding (Cohen's Kappa 0.705–1.0).
- High intra-class correlation coefficients (ICC) for both anatomical (0.908) and prosthetic (0.923) hands.
- Bland-Altman plots confirmed close scattering of data points, indicating validity.
Conclusions:
- Automating gaze data coding is feasible and effective.
- The developed algorithm significantly reduces coding time.
- Improved reliability and objectivity in analyzing gaze behavior for motor control.
