Real Time Gesture Recognition with Heuristic-Based Classification

Abstract

The recognition of human gestures in real time is still open problem due to low success rate of systems recently reported in scientific literature. This paper presents the proposed method and designed prototype for motion analysis and classification for human-computer interaction. The method is based on pattern recognition techniques of artificial vision without applying any markers or special sensors as well as utilizing low resolution cameras and simple hardware specifications. The proposed method provides interaction of user with computer via gestures in habitual and normal manner in order to activate system events (up, left and right) in real time. The proposed heuristic classifier recognizes specified gestures with an appropriate system context precision of 91.25 %. Comparing the obtained results with recent reports, the proposed approach provides satisfactory gesture recognition in real time with low resolution cameras.