A model for predicting pointing time in an eye-gaze input system using three basic phases of cursor movement trajectories.

Few attempts have been made to explain pointing time in an eye-gaze input system as a multi-part model for different movement phases. To improve the prediction accuracy of pointing time for an eye-gaze input system, we proposed a model to predict pointing time based on the observation of cursor movement trajectories from the start until the completion of pointing movement. The cursor movement process was classified into three phases: (i) latency response phase, (ii) ballistic eye movement phase, and (iii) homing eye movement phase. In phase (i), the relation was not affected by target size and movement distance, but was influenced by movement direction. The relation between the distance to a target center and the duration in phase (ii) was not affected by movement direction and target size. In phase (iii), the relation was affected by both target size and movement direction. Using the identified characteristics above, the model to predict the duration of each phase was proposed. Compared with the traditional Fitts' model, the proposed model resulted in an improved and satisfactory prediction accuracy of pointing time (contribution of more than 0.9) for an eye-gaze input system. The proposed model is expected to contribute to build up an effective and reliable model to predict pointing time necessary for the design and evaluation of human-computer interface (HCI) in an eye-gaze system.