Gesture localization in the test mode in the integral system of sign language training

Abstract

The article is devoted to the integral system of sign language training. The operation modes of the system based on 3D computer avatar are considered. The user’s feedback using the system is carried out in the test mode using a standard webcam. The user must demonstrate the word or dactyl specified by the system. The problem of gesture localization in the user test mode is considered. The algorithm for the rapid clustering of human hand gestures in a video sequence is proposed. The background removal is based on finding of the inter-element difference for the frames of a video sequence, the selection of the key frames and calculating difference frames. The use of the modified histogram method allows localizing the gesture. The experimental results show the stable operation of this algorithm in different light conditions. This approach allows using developed system both on stationary computers and on mobile devices.