Weighted Euclidean Distance Based Sign Language Recognition Using Shape Features

This paper proposes a real-time static hand gesture recognition system for American Sign Language alphabets. The input hand gestures from a simple background are captured by a camera and an image database is created. The proposed system consists of four stages namely preprocessing, segmentation, feature extraction, and classification. In the training phase, the hand region is detected and segmented from the gesture database images and various shape-based features such as area, perimeter, and roundness are extracted. The extracted features form a unique feature vector for a particular gesture. In the testing phase, the feature vector of an input test image is compared with each of the feature vectors of database images using weighted Euclidean distance. The gesture is correctly recognized if the distance is the least. This system is tested using a dataset of twenty-four ASL alphabets with three different signers. The experimental results show that the proposed system offers the recognition rate of 91.6 %.