Hand Gesture Recognition based on Invariant Features and Artifical Neural Network

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Objectives: Its objective is to develop a system which can recognize specific human hand gesture features from images and use these features to convey information for machine such as HCI (Human Computer Interaction). Method: this paper represents that, we propose a new method that supports the hand gesture recognition system in the static form, using SIFT feature extraction with feed forward neural network using MATLAB . We use SIFT technique to extract the invariant features of gesture sign. We have developed a gesture recognition system using feed forward neural networks which could recognize a finger alphabet of different types of symbols and where each gesture specifies a word. Findings: In the process of hand gesture recognition system there are many challenges addressed as: Illuminance conditions such that a small change in the lighting conditions effects badly on extraction process like color from which may produce misclassification problem. In the various scenes when the hand region rotates in any direction then the rotation problem get arises. When there is complicated backgrounds like there are other objects in the image with hand poses, this may refer to the background problems. Scale or Size problem, this problem may occur when the hand poses have not same size in the gestured image. At last the transformation problem, like from different different images it’s difficult to represent features of various hand positions. Improvements: To remove theses hindrances we have to develop an algorithm that can effectively recognize the hand gestures. Hand gestures is one of the techniques used in security purposes and robot communication.

This publication has 4 references indexed in Scilit: