Real time robust hand gesture recognition and visual servoing

The paper introduces hand gesture recognition and morphological model for visual servoing in real time unconstrained environments for positioning, docking or mobile target pursuing in robotics. Vision based hand tracking is an important problem in human computer interaction, since hand motions and gestures could potentially be used to interact with computers in more natural ways. A simple and fast process for gesture recognition and direction control is implemented. SIFT algorithm is used for hand gesture recognition by generating and comparing the feature vectors. Once the gesture is confirmed for particular system (e.g. motor) selection, further procedure would be to drive the system in required direction. For this purpose a morphological operation is performed and Hough transform is applied on the image of interest for angle measurement. This value is fed to the control system for further servo actions. The entire process is implemented in MATLAB. The paper includes experiments of 7 different hand postures and their results. Experiments show that the result is efficient and can be applied for real time applications.