A gesture-enabled method for natural identification in smart spaces

Smart spaces aim at providing personalized services to their inhabitants. This implies that the smart space has to be aware of who their users are, ideally through a seamless, informative and effortless procedure that should be integrated in the normal interaction model with the space. In this direction, this paper presents a method to naturally identify the user through his gestures in a smart environment. The system uses a Kinect sensor and relies on a supervised Dynamic Time Warping algorithm to process the position of the center of the user´s hand palm. In our tests, the system is able to identify with up to 30 gestures, which include directional and geometric movements (10) and small and capital letters (10-10). Results with 11 users and leave-one-out cross validation show that the system satisfactorily works when the user concatenates two gestures (letter-movement like compositions), providing an average correct classification rate (CCR) of 95.5% when the user freely chooses the gesture sentence. CCR may be up to 100% if the system itself proposes the identification sentence, including pairs of gestures that contain more “personal features” about the user.