05-05-2011, 03:48 PM
Abstract
Vision-based gesture recognition system isthe attractive solution for human computerinteraction and machine vision application likerobotic application. The common systemsassumption under such a systems are a constantenvironment, like persons wearing non-skin-coloredclothes with long sleeves and a fixed camera positionunder constant lighting conditions. In this paper weare evaluating the performance of a simple handand figure gesture recognition system that can bemapped to various tasks for robotic application.One of the main constrain for implementing thevision based gesture recognition system on roboticplatform is the computational complexity associatedwith the system because of the low computationpower available on the tiny embedded system onrobots. So here we are evaluating a simple handgestures for that purpose based on a simple heuristicalgorithm.
Index Terms—hand gesture, figure gesture, Heuristicalgorithm, machine vision, robotic application.
I. INTRODUCTION
Vision-based hand gesture recognition is an activearea of research in human-computer interaction (HCI),as direct use of hands is a natural means for humans tocommunicate with each other and more recently, withdevices in intelligent environments. The trend in HCI ismoving towards real-time hand gesture recognition andtracking for use in interacting with video games,remote-less control of television sets, robot interactionand interacting with other similar environments. Thissystem generally involves measurement of handlocation and shape. Achieving high accuracy and speedin measuring hand postures are two important aspectsof these systems.Numerous approaches have been explored toextract human hand regions either by backgroundsubtraction or skin-color segmentation [1, 2]. Methodsbased on background subtraction are not feasible whenapplied to images with complex backgrounds or realworldscenarios
II. SYSTEM MODEL
Fig.1 Hand Gesture Recognition system modelOur system uses very simple heuristic algorithmfor gesture recognition. Here hand portion of the imageis segmented by the simple algorithm sated as below insection IV and figure gesture recognition algorithm insection VI.
II. PATTERN RECOGNITION
Pattern recognition is the study of how machinescan observe the environment, learn to distinguishpatterns of interest from their background, and makesound and reasonable decisions about the categories ofthe patterns. Automatic (machine) recognition,description, classification, and grouping of patterns areimportant problems in a variety of engineering andscientific disciplines such as biology, psychology,medicine, marketing, computer vision, artificialintelligence, and remote sensing.The rapid growing and available compute power,while enabling faster processing of huge data sheets,has also facilitated the use of elaborate and diversemethods for data analysis and classification. At thesame time, demands on automatic pattern recognitionsystems are rising enormously due to the availability oflarge databases and stringent performance requirements(speed, accuracy, and cost).Here in the system verysimplest algorithm is used as pattern recognition.
Download full report
http://ijcnspdf/57.pdf