具頭動補償之高速可見光眼動儀系統平行架構設計

Abstract

眼動儀可應用於學習與認知心理學、商業廣告行為、神經科學等領域，利用眼睛觀看位置的資料進行統計分析，研究人類觀看行為的差異。目前市售的眼動儀多數使用紅外光技術，缺點為環境中的紅外光源會影響系統準確率，因此許多企業及學術單位投入於可見光眼動儀開發，但至今市面上尚未出現高精確度的可見光眼動儀產品。

本文以既有的眼球模型，改良虹膜抓取方法，針對一般辦公室光源環境下，使用每秒輸入480張影像之高速相機記錄眼睛影像。另一方面，利用高速相機影像連貫性的特徵，抓取眼睛固定特徵進行頭動補償，搭配平行化架構設計，以多執行序的技術，使本眼動儀計算速度可達到每秒480張影像，並達到高精確度與精準度的系統目標。Nowadays, the eye tracking system has been applied in the studies of learning and cognitive psychology, advertising design, neurosciences, and other fields. That is, the statistical analysis of the gaze data has made it possible to study the human behavior. Currently, most of gaze tracking systems are designed by equipped with an infra-ray (IR) light source, but the accuracy of an IR-based gaze tracking system can be affected by the illumination in the environment. Therefore, more and more researchers devoted themselves to the development of visible light eye tracking system. Still, few systems have been designed successfully which could reach a satisfactory level of accuracy and speed. The proposed system aims to improve the system accuracy as well as speed by modifying the iris matching algorithm. The breakthrough comes from the modifications of image preprocessing, head movement compensation, optimal iris matching, and a novel software architecture design based on parallel computing scheme. The experimental result shows that the proposed system has reached a process speed of 480 frames/s with a promising accuracy as well as precision result.