An Efficient Approach of Training Artificial Neural Network to Recognize Bengali Hand Sign
Alvi Mahadi, Fatema Tuj Johora, Mohammad Abu Yousuf · 2016
This work proposes a system that percepts hand signs and gestures via computer vision system and extract sufficient amount of images from it. After applying image processing and extracting the features of the images, the system uses an algorithm to recognize the hand signs and gestures. In the process of recognizing the hand signs, the Artificial Neural Network (ANN) is being trained with some specific data sets obtained from the feature extraction of the images. Then the system applies Instance Based Learning with a back-propagated neural network and Pattern matching Techniques to make the Artificial Neural Network learn and classify the signs and gestures. This approach of recognition decreases the error rate which increases the efficiency of the system.