Static Hand Gesture Recognition

Abstract

We have presented two hand gesture recognition methods, namely, edge detection based recognition and correlation based recognition of images of hand gestures of American Sign Language (ASL) in a constrained environment. Both above mentioned methods are imagecomparison method, in which various parameters are compared, namely, edge count in edge detection based recognition and correlation coefficient in correlation based recognition. Edge detection based method was motivated from image-subtraction method and edges were detected based on sobel edge detection method. Test image is compared to all the images of corresponding letters of ASL and gesture are recognised based on the percentage of match of edge count in edge detection based recognition and correlation coefficient in correlation based recognition, among the database and test images. To evaluate the performance of the algorithms it is tested on 4200 images for edge detection method and 260 for correlation method. Results show that excluding some similar hand shapes and testing in constrained environments, accuracy is more than 90 percent for both the methods in a constrained environment.