Vision based human interaction system for disabled

This research work is related to the application of machine vision technique to develop a robust assistive human computer interaction technology for those with physical accessibility problem of controlling mouse and keyboard with hand. Paper's main motif is inferring information about planer movement of the head using a video camera and transforming this motion to the pixel coordinate system of the display so as to control the position of mouse pointer. Iterative sparse optical flow algorithm computes the pattern of apparent motion between sequential facial image frames captured by the webcam. Adaboost based Cascaded Harr classifier is used to detect face and eye across frames and we have given special attention towards the issues involving drawback regarding misdetection of tilted faces in the image frame inspite of training our datasets with tilted facial images. Left/Right eye blink is used to control the clicking event of mouse. Blink of eye is modelled by fitting the trained data using Spline and Gaussian curve which determines the likelihood function to determine the posterior probability. This work is motivated by the need to design an affordable real-time system in the interest of a large serving community.