Hand Gesture Recognition Using Neural Networks For Robotic Arm Control

Abstract

Hand gesture recognition is one of well-research vision system projects. There are numerous objects which use different techniques to tackle different aspect of the problem. In the paper, we present a project to use hand gesture recognition system for controlling a 5 degreeof freedom robotic arm, i.e. SCORBOT. We use cam-shift Algorithm, Principal Component Analys (PCA) and Artificial Neural Networks (ANNs) to build the image-based hand gesture recognition system. Cam-shift is used to track the hand image and PCA is use to reduce the feature dimensions of image. At the end of the recognition system, we employ Artificial Neural Network to classify the image of static hand gesture. We have successfully developed natural language to control the robot and our system is able to recognize six different static hand gesture (proses), handle some noise of image acquistous process, and implement some modes of robot control. The average performance of the system to recognize the static hand gestures is larger than 90% and the robot is able to do some simple jobs by using hand gesture language commands as its input.