Applications of motion and its biological representation in recognition of gestures

In this article, the problem of gesture recognition is approached using motion information. Motion contains important information that is used by both people and animals for perception of their surroundings. In this work, a web camera is used to capture the images from which motion information is extracted. The model obtained will be used in the future in a demonstration learning system applied to robotics. To approach this problem, gesture recognition has been identified as the first step, and for this purpose, the architecture requires that three main aspects be solved: the instantaneous representation of motion, the integration of this information into time, and the strategy of classification. In this work, only the first of them is considered. In contrast to previous works, in this paper, the extraction of motion and its codification are inspired in the motion processing made by the brain of macaques. The model shown has been applied to the recognition of four gestures made by different people for which the percentage of gesture recognition was 89.6% with a robustness of 96.9% from different points of view.