Virtual Mouse for Amputee using Computer Vision and CNN-SVM

As digital technology becomes increasingly integrated into everyday life, accessibility remains a significant challenge for individuals with physical disabilities, particularly amputees. This paper presents a novel virtual mouse system that enables hands-free computer interaction through eye movement detection using computer vision techniques. The proposed system leverages a hybrid model combining convolutional neural networks (CNNs) and support vector machines (SVMs) to accurately recognize gaze direction and blinking patterns, enabling core mouse operations such as clicking and scrolling. Unlike previous approaches that rely on specialized hardware, our method is cost-effective and practical, requiring only a standard camera for operation. Experimental evaluation across 300 trials demonstrates high precision and recall, especially in detecting left and right click actions. The results highlight the system’s reliability, usability, and potential to significantly improve digital accessibility for users with motor impairments.