• Version
• Download 15
• File Size 590.83 KB
• File Count 1
• Create Date 20 March 2026
• Last Updated 20 March 2026

EyeFlow: A Web-Enabled Eye-Tracking Mouse for Hands-Free Computer Interaction

D. Parthiban Department of CSE

Dr. M.G.R Educational and Research Institute Maduravoyal, Chennai parthiban.parthi2904@gmail.com

D. Mohammed Rafeek Department of CSE

Dr. M.G.R Educational and Research Institute Maduravoyal, Chennai mdrafeek1911@gmail.com

Mohammed Fahij Department of CSE

Dr. M.G.R Educational and Research Institute Maduravoyal, Chennai Yunusnoor80@gmail.com

Dr. A. Jegadeeswari, Assistant Professor Department of Computer Science and Engineering Dr. M.G.R Educational and Research Institute Maduravoyal, Chennai jegadeeswari.cse@drmgrdu.ac.in

Dr. T. Kumanan, Professor

Department of Computer Science and Engineering Dr. M.G.R Educational and Research Institute Maduravoyal, Chennai kumanan.cse@drmgrdu.ac.in

Dr. M. Nisha, Assistant Professor

Department of Computer Science and Engineering Dr. M.G.R Educational and Research Institute Maduravoyal, Chennai

nisha.cse@drmgrdu.ac.in

Abstract

Typical keyboard-and-mouse input needs some fine motor control that can hardly be consistently maintained by some 1.3 billion individuals on earth. One method to overcome this obstacle is an eye-tracking device specifically designed to do this, though even a relatively inexpensive consumer product, such as the Tobii Eye Tracker 5, costs about 229, and research-grade a device will cost between 5,000 and 15,000, which is out of the reach of many potential customers. The present paper describes an eye-tracking- based flow control system, EyeFlow (an eye-tracking flow control system) that is a software-only assistive interface that operates on a standard webcam. It was four-channel interactive: (1) head- compensated gaze-based cursor control (using relative movement between the iris and nose tip) to calculate mouse clicks, (2) right- eye mouse clicking using the Eye Aspect Ratio (EAR) algorithm,

(3) nose-tilt page scrolling using asymmetric gain factors, and (4) continuous voice typing using Google Speech Recognition. The Face Mesh MediaPipe [2] can extract 468 facial landmarks at 30 fps; A five-point automatic calibration then distorts the motion of the iris to screen coordinates and no keyboard input is required. In a study with 10 participants in three sessions, the mean error of placement of the cursor was 3.8 percent of the screen diagonal ( =.9 percent ), wink-detection was 95.2 percent, and the end-to-end latency of a cursor was 62 ms ( =14 ms). Strongness testing ensured that the error of head motion was less than 8% in the depth and 15 cm in the lateral direction. EyeFlow can run on a regular laptop, and requires no extra hardware to be purchased (i.e. no additional costs are incurred of its own). At about 95% cost reduction over entry-level commercial systems (e.g. Tobii Eye Tracker 5 at 229 vs. 0 incremental cost of EyeFlow), it can run on a regular laptop with no extra hardware. The processing of all videos was done on-site, and there was no facial data transmission outside of it; it was only the speech typing part that was connected to the Google Speech Recognition cloud service.

Index Terms—Eye tracking, gaze estimation, assistive technology, blink detection, Eye Aspect Ratio, voice typing, MediaPipe, head- compensated gaze tracking, webcam-based interaction, accessible computing.