Real time Hand Gesture Recognition using different algorithms based on American Sign Language

Abstract

Human Computer Interaction (HCI) is a broad research field based on human interaction with computers or machines. Basically, Hand Gesture Recognition (HGR) is a subfield of HCI. Today, many researchers are working on different HGR applications like game controlling, robot control, smart home system, medical services etc. The purpose of this paper is to represent a real time HGR system based on American Sign Language (ASL) recognition with greater accuracy. This system acquires gesture images of ASL with black background from mobile video camera for feature extraction. In the processing phase, the system extracts five features such as fingertip finder, eccentricity, elongatedness, pixel segmentation and rotation. For feature extraction, a new algorithm is proposed which basically combines K curvature and convex hull algorithms. It can be called “K convex hull” method which can detect fingertip with high accuracy. In our system, Artificial Neural Network (ANN) is used with feed forward, back propagation algorithm for training a network using 30 feature vectors to recognize 37 signs of American alphabets and numbers properly which is helpful for HCI system. The total gesture recognition rate of this system is 94.32% in real time environment.