The Design And Implementation Of Eye Movement-Based Text Input System

When people interact with computers, the sight of individual’s eyes often contains current task condition and users internal state. Thus, it is worth studying how to apply eye tracking technology to the field of human computer interaction. Because of the high ambiguity and low accuracy of eye behaviors, the eye movement-based interaction is still a challenging task.This thesis begins by summarizing the advantages and limitations of the existing eye movement-based interaction system. Then, based on the "user centered design" principle, a model of eye movement-based interaction system (MEMIS) aimed at promoting the system usability and user experience is proposed. This model contains four layers, activity layer, data layer, behavior layer and application layer. In the behavior layer, an eye behavior real-time recognition algorithm based on the fixation model and the saccade model is presented, which meets the interaction requirements of different people. Next, the eye input mechanism is defined and experiment are conducted to solve problems of this mechanism.Finally, according to the design principle of eye movement-based interaction, a text input system based on eye input is designed and implemented. In this system’s design, the basic interface and the saccadic high-level function are constant, while the dwell time parameter, the feedback mode and character prediction mechanism are variable. To optimize the design, four contrast experiments are performed and the system is evaluated from the perspective of system usability and user experience. Evaluation results show that under the condition of optimal dwell time threshold, optimal feedback mode and opening character prediction mechanism, user’s text input speed reaches 20 words per minute, which meets the requirements of common interactions. Furthermore, when people uses this system for a long time, people will learn the eye input modal, and people’s work load decrease.