Computer vision based framework for digit recognition by hand gesture analysis

In recent years, hand gesture recognition framework as an effective sign language tool has been extensively explored by many researchers. This paper presents an idea of developing a framework using computer vision for hand gesture based sign language recognition from real-time video stream. The proposed system identifies hand-palm in video stream based on skin color and background subtraction scheme as opposed to the conventional techniques of using gloves or markers as interface and thereby makes an effort of exploring the possibility of a suitable computer vision framework for hand gesture recognition. We have also proposed an iterative polygonal shape approximation strategy in fusion with a special chain-coding scheme for shape-similarity matching. This proposed framework considers digits as important symbols of sign language and it successfully recognizes hand gestures corresponding to various numerical digits with acceptable accuracy.