New Cyclical Pattern and Temporal-Spatial Representation for Robust Dynamic Hand Gesture Recognition

This paper tackles a new prototype of dynamic hand gestures and its advantages to apply in controlling smart home appliances. The proposed gestures convey cyclical patterns of hand shapes as well hand movements. Thanks to the periodicity of defined gestures, on one hand, some common technical issues that may appear when deploying the application (e.g., spotting gestures from a video stream) are addressed. On the other hand, they are supportive features for deploying robust recognition schemes. To this end, we propose a novel hand representation in temporal-spatial space. Particularly, the phase continuity of the gesture’s trajectory is taken into account underlying the conducted space. This scheme obtains very competitive results with the best accuracy rate is of 96%. We deploy the proposed techniques to control home-appliances such as lamps, fans. Such systems have been evaluated in both lab-based environment and real exhibitions. In the future, the proposed gestures will be evaluated in term of naturalness of end-users and/or robustness of the systems.