DESIGN AND IMPLEMENT A SYSTEM OF GRASP IDENTIFICATION FOR DEXTEROUS ROBOT HAND

Abstract

The grasp taxonomy is one of the key problems in grasp planning of dexterous robot hands. This paper describes the design and implement of a system of grasp identification based on pattern recognition technology. The feature of human grasp is represented by Gaussian Mixture Model (GMM). The parameters of GMM are estimated by the Expectation Maximization algorithm (EM). The taxonomy of robot hand grasps is achieved by mapping grasps from human hand to robot hand. The effectiveness of the GMM identification method is proved by the experiments. The system can be used in grasp planning.