Abstract
We present a prototype system for secure human-machine interaction, based on face and arm gesture recognition. A mobile robot mounted with two CCD cameras captures the scene continuously, and gets ready to follow the instructions, if it finds a familiar face in vicinity. Face matching is done by the elastic graph matching using the Gabor wavelet transform coefficients. Calculation of the coefficients and the matching algorithm are both implemented in hardware for better computational efficiency. The robot responds only to the commands from the authorized users. The commands are passed by arm gesture, recognized by the robot using a stereo vision based algorithm.