Real Time Hand Gesture Recognition and Blink Detection

Real Time Hand Gesture Recognition and Blink Detection

Abstract

This was one of the first projects I implemented in computer vision, without much knowledge of machine learning or advanced computer vision techniques. The objective of this project was to implement the mouse tasks of cursor movement, left click and right click in a Windows OS using hand gestures. A simple webcam was used for capturing frames continuously. Hand region was found in real time by background subtraction and color segmentation in HSV color space. Hand features, which include fingertips and thumb were detected using contours. Fingertip count, position and thumb detection were used to form three distinct hand gestures.

As an extension, with intention to control cursor using eyes, intentional blink detection was implemented using eye region extraction, histogram equalization and blob detection. Natural blinks were ignored. This feature was not used for implementing mouse tasks.