Multimodal, Embodied and Location-Aware Interaction

This work demonstrates the development of mobile, location-aware, eyes-free applications which utilise multiple sensors to provide a continuous, rich and embodied interaction. We bring together ideas from the fields of gesture recognition, continuous multimodal interaction, probability theory and audio interfaces to design and develop location-aware applications and embodied interaction in both a small-scale, egocentric body-based case and a large-scale, exocentric `world-based' case. BodySpace is a gesture-based application, which utilises multiple sensors and pattern recognition enabling the human body to be used as the interface for an application. As an example, we describe the development of a gesture controlled music player, which functions by placing the device at different parts of the body. We describe a new approach to the segmentation and recognition of gestures for this kind of application and show how simulated physical model-based interaction techniques and the use of real world constraints can shape the gestural interaction. GpsTunes is a mobile, multimodal navigation system equipped with inertial control that enables users to actively explore and navigate through an area in an augmented physical space, incorporating and displaying uncertainty resulting from inaccurate sensing and unknown user intention. The system propagates uncertainty appropriately via Monte Carlo sampling and output is displayed both visually and in audio, with audio rendered via granular synthesis. We demonstrate the use of uncertain prediction in the real world and show that appropriate display of the full distribution of potential future user positions with respect to sites-of-interest can improve the quality of interaction over a simplistic interpretation of the sensed data. We show that this system enables eyes-free navigation around set trajectories or paths unfamiliar to the user for varying trajectory width and context. We demon- strate the possibility to create a simulated model of user behaviour, which may be used to gain an insight into the user behaviour observed in our field trials. The extension of this application to provide a general mechanism for highly interactive context aware applications via density exploration is also presented. AirMessages is an example application enabling users to take an embodied approach to scanning a local area to find messages left in their virtual environment.