Real-time non-intrusive eye-gaze tracking based wheelchair control for the physically challenged

Driving and control of wheelchair by hand poses a great impediment for a physically challenged person with limited physical ability but a real time eye-gaze tracking based control system can go a long way in easing such cases. A robust scheme of pupil movement detection based on Recursive Circular Hough Transform (RCHT) is proposed in this paper. Despite CHT being a common technique for pupil detection, the requirement of a pre-defined contrast sensitivity and radius range of the target circular object limits its use in real time applications where both sensitivity and radius of the pupil may vary widely. Moreover, in different gazing modes, eyes may not appear to be circular either. Again, edge sensitivity reduction and relaxation of circular object detection criterion causes CHT to detect multiple objects other than just pupil in a given contrast sensitivity. To eliminate such problems, an RCHT technique is proposed which is able to automatically tune the sensitivity and radius. Thus, a robust performance without the usage of computationally expensive classifiers can be obtained. Very low resolution off-the-shelf cell phone camera has been used to capture images, from where pupil location detection has been performed. Effects of various changing conditions such as pupil color, camera position, speed of eye movement etc. have been analyzed too. Results assert that this computationally light approach provides satisfactory accuracy on the basis of correctly following the actual instructed direction in test cases and thus, it is suitable for real-time precise direction control for people with working eyes but very constrained physical abilities.