AI-Based Virtual Mouse Controller Using Computer Vision
Dr. Shaik Irfan Babu,
Assistant Professor
Department of Emerging Technologies, Mahatma Gandhi Institute of Technology Gandipet, Hyderabad, Telangana, India skirfanbabu_cse@mgit.ac.in
Mrs. A. Swapna,
Assistant Professor
Department of Emerging Technologies, Mahatma Gandhi Institute of Technology Gandipet, Hyderabad, Telangana, India aswapna_cse@mgit.ac.in
Bhanu Teja,
Student
Department of Emerging Technologies, Mahatma Gandhi Institute of Technology Gandipet, Hyderabad, Telangana, India [kbhanuteja_csm226625@mgit.ac.in]
Parameshwar,
Student
Department of Emerging Technologies, Mahatma Gandhi Institute of Technology Gandipet, Hyderabad, Telangana, India [kparameshwarreddy_csm226629@mgit.ac.in]
ABSTRACT
The AI-Based Virtual Mouse Controller using Computer Vision provides a touch-free method for controlling a computer through simple hand gestures captured by a webcam. This system eliminates the need for a physical mouse, offering a hygienic, accessible, and intuitive human–computer interaction method, especially useful for general users as well as individuals with motor impairments. By leveraging advancements in artificial intelligence and real-time hand-tracking technology, the system interprets natural hand movements to perform essential mouse operations such as cursor navigation, clicking, and dragging, enabling smooth and responsive touchless computing. The system operates by capturing live video frames through OpenCV and processing them using MediaPipe's hand-landmark model, which extracts 21 key hand points for gesture analysis. Based on the relative positions of these landmarks, gestures such as raising the index finger for cursor movement or bringing fingers together for clicking are recognized and mapped to OS-level mouse actions using PyAutoGUI. Smoothing techniques are applied to reduce jitter and ensure stable cursor motion, resulting in an efficient, accurate, and fully functional virtual mouse interface suitable for real-time practical applications.The system demonstrates stable real-time performance with low latency and consistent accuracy under standard indoor conditions without requiring any training dataset or specialized hardware.
Index Terms – Hand Gesture Recognition, Virtual Mouse, Computer Vision, Human-Computer Interaction, OpenCV, Contactless Interface
