ADVANCED GESTURE-BASED VIRTUAL MOUSE USING MATHEMATICAL LANDMARK MAPPING

This paper presents an Advanced Gesture-Based Virtual Mouse system that enables touchless, real-time control of a computer cursor through hand gestures captured by a standard webcam. The proposed system leverages Googles Media Pipe framework for robust skeletal hand landmark detection and OpenCV for efficient video frame processing. Specifically, the system tracks twenty-one anatomical hand landmarks per frame, using Landmark 8 (index fingertip) for cursor movement and computing the Euclidean distance between Landmark 4 (thumb tip) and Landmark 8 to detect a pinch-click gesture. A bilinear affine coordinate transformation maps the index fingertips pixel position within a predefined active camera zone to full screen coordinates, and an exponential weighted smoothing algorithm suppresses jitter to ensure fluid cursor motion. The click mechanism employs a real -time Euclide an distance threshold, triggering a mouse click event only when the inter-landmark distance falls below 35 pixels, with a 200 ms debounce interval to eliminate accidental double-clicks. Implemented in Python using the PyAutoGUI automation library, the system operates at approximately 28-32 frames per second on standard consumer hardware, achieving a click detection accuracy exceeding 94% under controlled lighting conditions.