Abstract: Previous eye trackers had complex structure,huge size and weight,and are only desktop oriented.To support ubiquitous computing and mobile interaction,an approach for eye tracking based on pupil center cornea reflection is introduced.The approach includes image processing,features detection,data processing and interaction application.Image filtering and binarization are used first.And then the improved algorithm of ellipse fitting and template matching is used to detect pupil and Purkinje images respectively.Finally,eye-movement data can be calculated,such as fixation's coordinates.A head mounted eye tracker prototype is presented,and the results of user study indicates it is not uncomfortable for users,and has high accuracy and robustness.It furthermore demonstrates the feasibility and effectiveness for the eye tracking based mobile interaction.