Séminaire de Mme Doan Thi Huong Giang, doctorante de l'Institut MICA - Date : le 27 avril 2017, 14h00 - Lieu : seminar room, Institut MICA, Hanoi University of Science and Technology
Intervenant :
Mme Doan Thi Huong Giang, doctorante du Département Computer Vision de l'Institut MICA
Date : jeudi 27 avril 2017, 14h00
Lieu : salle "seminar room", 9ème étage, bâtiment B1, Institut MICA, Hanoi University of Science and Technology
Langue : le séminaire sera présenté en anglais
Résumé/Abstract:
Human-Computer Interaction using hand gestures has widely increased in recent years. However, this type of research faces many technical issues such as real-time requirement and complexity of hand movements. In this talk, a real-time and accurate hand gesture recognition for automatic controlling electronic appliances in a smart-home such as lighting, fans are presented. We demonstrated that it is a natural and friendly way replacing the conventional remote controller. In order to obtain this, a unified solution to address the different issues is proposed. Firstly, we tackled a new set of dynamic hand gestures and its advantages solve critical issues that may appear when deploying the real application. The proposed gestures convey closed-form patterns of hand shapes as well hand movements. Utilizing these characteristics, hand regions are segmented and the sequence of hand gestures are extracted from the video stream. For gesture recognition, we proposed a novel temporal-spatial space which takes into account combination of the trajectories of hand movements and the low non-linear reduction space. Particularly, the phase continuity of the gesture’s trajectory is paid much attention underlying the conducted space. We show that phase's registration is an important factor that affects the system’s performance. Finally, the system is tested in a real application to control home appliances such as lamps, fans in both lab-based environment and real exhibitions. End-users’ behaviors and robustness of the application is reported. A plan for the future is to improve hand gesture recognition including features to be used and algorithms.