A Cost-Effective Webcam Eye-Tracking Algorithm for Robust Classification of Fixations and Saccades

Dementia, particularly Alzheimer's disease (AD), poses significant challenges to cognitive function. Utilizing eye-tracking devices for early AD detection has shown promise, but expensive lab-grade equipment limits accessibility. This paper extends previous work by proposing a robust eye movement classification system using a cost-effective webcam eye-tracker to measure fixations and saccades. The algorithm is evaluated against established methods on Tobii and webcam data, exhibits competitive accuracy, especially excelling in saccade identification. Beyond its efficacy in Alzheimer's research, the proposed algorithm demonstrates versatility and reliability, positioning it as a valuable tool for various applications requiring precise eye movement characterization. The study utilizes a dataset collected from 23 participants without known cognitive decline, comprising 27 samples. This paper evaluates established eye movement classification methods alongside the novel algorithm using both lab-grade Tobii eye-tracker and webcam eye-tracker data. The proposed method led to F1 = 0.93 in detecting fixations, comparable to a Random Forest method (F1 = 0.95). However, the new method also led to considerable improvements in saccade classification, with F1 = 0.69, compared to the next best method with F1 = 0.40.