Real-Time Precise Human-Computer Interaction System Based on Gaze Estimation and Tracking

Table 5

Comparison results.

Literatures	How to interact	Key methods	Accuracy	Speed	Comparison conclusions
[18]	The screen is divided into , which can monitor the user’s gaze on the screen in real time and can be used as 9 or 25 interactive channels	Improved pupil detection algorithm based on color intensity	: 94%; : 78%	In real time	It had poor accuracy and insufficient flexibility for practical applications.
[24]	Use the nine key inputs widely applied in straight board mobile phones	Using CNN to estimate the gaze in nine directions, representing nine keys	90%	20 letters per minute for a skilled user	It was twice as fast compared to our work, but it had insufficient flexibility and limited functions.
[26]	Move the mouse through the head posture, and control the mouse switch, click and scroll through blinking and other facial expressions	Head posture estimation, face detection, and expression recognition	Not mentioned	Not mentioned	It requires various operations to adjust the position of the head when moving the mouse, which reduces the interaction efficiency and usability.