研究目的
Investigating the possibility of detecting task demand automatically, reliably, and unobtrusively using eye movements only to design advanced decision support systems.
研究成果
The study demonstrated that task demand can be detected automatically, reliably, and unobtrusively via eye movements. Pupil data, especially the ratio of pupil dilation during saccades and fixations, was the most important predictor of task demand. The results support the feasibility of building advanced task load detection systems using eye tracking and machine learning.
研究不足
The study is limited by the task context (math problem-solving), a relatively static population (graduate students in an engineering school), and the manipulation of task demand via time limit only. Future studies could explore different tasks, more diverse populations, and other task characteristics to extend the results.
1：Experimental Design and Method Selection:
The study involved developing an eye tracking task load detection system to classify eye movements under different task demands. The methodology included the use of a Random Forest algorithm for classification.
2：Sample Selection and Data Sources:
Participants were graduate students from technical disciplines, randomly assigned to control or experimental groups. The experimental group had a time limit for task completion, while the control group did not.
3：List of Experimental Equipment and Materials:
Tobii X300 remote eye tracker with a sampling rate of 300 Hz mounted on a 21-inch monitor at a resolution of 1920 x
4：Experimental Procedures and Operational Workflow:
12 Participants completed a set of mathematical questions under two different task treatments (with and without time limit). Eye movements were recorded during the task.
5：Data Analysis Methods:
Eye movement data was analyzed using the Random Forest algorithm to classify task demand. The performance of the classifier was assessed using accuracy, confusion matrix, and ROC curve.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
