Image classification for sign language usingconvolutional neural network

ABSTRACT MSC THESIS IMAGE CLASSİFİCATİON FOR SİGN LANGUAGE USİNG CONVOLUTİONAL NEURAL NETWORK KHALED MOHAMED ABUBAKER ELBAYOUDI KASTAMONU UNIVERSITY INSTITUTE OF SCIENCE DEPAERTMENT OF MATERİAL SCİENCE AND ENGİNEERİNG SUPERVISOR: ASSİST. PROF. ÜMİT TOKEŞER In this thesis, the deep learning method is used to Turkish Sign Languages. Many studies on sign language recognition have been and are still being conducted around the world. When the studies on sign language recognition were evaluated, it was seen that many problems were encountered. From the perspective of image processing, the main problems are the light in the environment, the complexity of the background, the position of the camera, the difficult and troublesome removal of the relevant region, the loss of some signs as a result of segmentation applications and the movements of the finger alphabet are very similar. There are also problems such as incorrect results when the signs are not made in accordance with the rule, difficulty in selecting the right frames for the recognition of successive signs, and inability to choose the correct property extraction method. When we look at the latest studies, it is seen that Kinect technology is used to solve the problems caused by image processing. Such systems are very costly as they require additional equipment and are therefore not preferred much. Therefore, studies are still going on to develop sign language recognition systems with the least error rate. KEYWORDS: Turkish sign languages, artificial neural network, deep learning. 2020, 69 Pages Science Code:91