Vision-based recognition of gestures with context

摘要: Out of the large range human actions, gestures performed with hands play a very important role during everyday life. Therefore, their automatic recognition is highly relevant for constructing userfriendly human-machine interfaces. This dissertation presents new approach to manipulative that interact objects in environment. The proposed gesture can serve realize pro-active interfaces enabling technical systems observe humans acting environment and react appropriately. For observing motions natural hands, non-intrusive vision-based technique required. For this purpose, an adaptive skin color segmentation capable detecting skin-colored wide lighting conditions developed. adaptation step controlled by using additional scene information restrict updating model image areas actually contain areas. With trajectory data results from sequence images, be performed. To recognize context, method incorporating process described. context consists current state hand object manipulated. needed capture applicability while manipulated needs present vicinity enable recognizing model. Through integration, developed system allows are mainly characterized interaction do not have characteristic trajectory. performance demonstrated assembly construction scenario typical office The use improving shown applying 'situated artifical communicator' situated domain. Here about executed used dialog providing contents. Besides direct improvement interface, recognized also as knowledge other components. observation provide vision algorithms aiming at assemblies scenario. In way improve interface indirectly.