Real-Time Eye-Movement-Based Mouse Cursor Control Using VGG16 with Mouth-Opening Ratio for Click and Scroll Actions

Managing computer interfaces without conventional input tools provides considerable advantages for people with motor disabilities and facilitates innovative interaction methods.This research presents a real-time mouse cursor control system driven by eye movement and enhanced with click and scroll functionalities using facial gestures.The proposed method employs a VGG16-based convolutional neural network to estimate the user's gaze position from webcam images.Dual eye-region inputs are processed to predict on-screen cursor coordinates, enabling smooth and accurate pointer movement.Click and scroll commands are implemented through a non-intrusive mouth-opening ratio (MAR) metric, calculated from facial landmarks.Short blinks or predefined facial cues trigger left and right clicks, while varying mouth-opening levels initiate scroll up and scroll down actions.A custom dataset, collected under varied lighting and user conditions, is combined with public gaze datasets for robust model training and fine-tuning.The system operates in real time, integrating prediction outputs with operating system mouse events for seamless control.Experimental evaluation reports competitive gaze estimation accuracy, high click detection precision, and reliable scroll triggering, even in unconstrained environments.This work demonstrates the feasibility of combining deep learning-based gaze tracking with intuitive facial gesture controls, paving the way for accessible, hands-free human-computer interaction solutions.