PARAMETER CALIBRATION FOR RECONFIGURATION OF A HAND GESTURE TELE-ROBOTIC CONTROL SYSTEM

Abstract

In this work we design two versions of a local neighborhood search algorithm. These versions are customized for a system operational parameter calibration task. The first and second methods perform complete and incomplete probabilistic neighborhood searches, respectively. The domain of application of the method is that of hand gesture control of telerobotic systems. The primary need for recalibrations of such systems is frequent relocation to other environments such as; laboratories, and remote control stations. A secondary need for recalibration, is the custom redesign of the gesture control language. This occurs for new users, new control tasks and new vocabularies. Allowing for fast recalibration of system parameters provides the system flexibility to respond to such new system setups. The two proposed methods were compared using a test case of 13 gesture commands. Both provided hand gesture recognition accuracies of 99.78. However, the probabilistic version had 48.5 percent less solution evaluations. Needless to say, the methodology reported here has wider application. Index Terms — fuzzy c-means, gesture recognition, hand gestures, neighborhood search, telerobotics, supervised clustering, automated setup. parameter calibration, reconfiguration.