A strong driving force of scientific progress in the technical sciences is the quest for systems that assist humans in their daily life and make their life easier and more enjoyable. Nowadays smartphones are probably the most typical instances of such systems. Another class of systems that is getting increasing attention are intelligent robots. Instead of offering a smartphone touch screen to select actions, these systems are intended to offer a more natural human-machine interface to their users. Out of the large range of actions performed by humans, gestures performed with the hands play a very important role especially when humans interact with their direct surrounding like, e.g., pointing to an object or manipulating it. Consequently, a robot has to understand such gestures to offer an intuitive interface. Gestural understanding is, therefore, a key capability on the way to intelligent robots.

This book deals with vision-based approaches for gestural understanding. Over the past two decades, this has been an intensive field of research which has resulted in a variety of algorithms to analyze human hand motions. Following a categorization of different gesture types and a review of other sensing techniques, the design of vision systems that achieve hand gesture understanding for intelligent robots is analyzed. For each of the individual algorithmic steps – hand detection, hand tracking, and trajectory-based gesture recognition – a separate Chapter introduces common techniques and algorithms and provides example methods. The resulting recognition algorithms are considering gestures in isolation and are often not sufficient for interacting with a robot who can only understand such gestures when incorporating the context like, e.g., what object was pointed at or manipulated.

Going beyond a purely trajectory-based gesture recognition by incorporating context is an important prerequisite to achieve gesture understanding and is addressed explicitly in a separate Chapter of this book. Two types of context, user-provided context and situational context, are reviewed and existing approaches to incorporate context for gestural understanding are reviewed. Example approaches for both context types provide a deeper algorithmic insight into this field of research. An overview of recent robots capable of gesture recognition and understanding summarizes the currently realized human-robot interaction quality.

The approaches for gesture understanding covered in this book are manually designed while humans learn to recognize gestures automatically during growing up. Promising research targeted at analyzing developmental learning in children in order to mimic this capability in technical systems is highlighted in the last Chapter completing this book as this research direction may be highly influential for creating future gesture understanding systems.