Dynamic Gesture Recognition—A Machine Vision Based Approach

Abstract

Computationally simple method for dynamic hand gesture recognition is presented in this paper. The segmentation of hand which take parts in gesture production is being addressed tow different ways using color band based segmentation algorithms. The first one uses the special color stickers as part of finger and the second one does segmentation based on normal skin color. The movement of hand is tracked and Freeman’s eight directional code is generated corresponds to each gestures. A dynamic time wrapping based Levenshtein minimum edit distance algorithm is used for classification. The results of dynamic hand gestures with special color approach and without special color are discussed separately. The accuracy of the system is found to be more for special colour based segmentation than skin colour based segmentation techniques.