ANN-Based Hand Gesture Recognition Using Self co-articulated Set of Features

Abstract

ABSTRACTHand gesture recognition system can be a particularly challenging task because of the presence of several spatio-temporal variations like hand trembling, self co-articulation within a gesture, etc. during the segmentation stage. In this paper, we focus our attention on self co-articulation problem which has not been addressed before. With the addition of a new feature, i.e. removing the hand movement during pause state along with the velocity features within the gestures, we were able to perform the gesture spotting more smoothly for the isolated gestures. Moreover, a new set of novel features was added in the feature extraction stage: (1) number of self co-articulated stroke; (2) orientation of the self co-articulated strokes; (3) position of the hand; (4) distance between the start and end points of a gesture; and (5) ratio between the longest and shortest distances from centre. These features were fed to neural network classifier. The results of experimentation were carried out with 40 gestures and it showed an average true positive rate of 91.25%.