Supporting a wide set of linked non-verbal resources remains an evergreen challenge for communication technology, limiting effectiveness in many applications. Interpersonal distance, gaze, posture and facial expression, are interpreted together to manage and add meaning to most conversations. Yet today's technologies favor some above others. This induces confusion in conversations, and is believed to limit both feelings of togetherness and trust, and growth of empathy and rapport. Solving this problem will allow technologies to support most rather than a few interactional scenarios. It is likely to benefit teamwork and team cohesion, distributed decision-making and health and wellbeing applications such as tele-therapy, tele-consultation, and isolation. We introduce withyou, our telepresence research platform. This paper describes the end-to-end system including the psychology of human interaction and how this drives requirements throughout the design and implementation. Our technology approach is to combine the winning characteristics of video conferencing and immersive collaborative virtual environments. This is to allow, for example, people walking past each other to exchange a glance and smile. A systematic explanation of the theory brings together the linked nature of non-verbal communication and how it is influenced by technology. This leads to functional requirements for telepresence, in terms of the balance of visual, spatial and temporal qualities. The first end-to-end description of withyou describes all major processes and the display and capture environment. An unprecedented characterization of our approach is given in terms of the above qualities and what influences them. This leads to non-functional requirements in terms of number and place of cameras and the avoidance of resultant bottlenecks. Proposals are given for improved distribution of processes across networks, computers, and multi-core CPU and GPU. Simple conservative estimation shows that both approaches should meet our requirements. One is implemented and shown to meet minimum and come close to desirable requirements.

The aim of our experiment is to determine if eye-gaze can be estimated from a virtuality human: to within the accuracies that underpin social interaction; and reliably across gaze poses and camera arrangements likely in every day settings. The scene is set by explaining why Immersive Virtuality Telepresence has the potential to meet the grand challenge of faithfully communicating both the appearance and the focus of attention of a remote human participant within a shared 3D computer-supported context. Within the experiment n=22 participants rotated static 3D virtuality humans, reconstructed from surround images, until they felt most looked at. The dependent variable was absolute angular error, which was compared to that underpinning social gaze behaviour in the natural world. Independent variables were 1) relative orientations of eye, head and body of captured subject; and 2) subset of cameras used to texture the form. Analysis looked for statistical and practical significance and qualitative corroborating evidence. The analysed results tell us much about the importance and detail of the relationship between gaze pose, method of video based reconstruction, and camera arrangement. They tell us that virtuality can reproduce gaze to an accuracy useful in social interaction, but with the adopted method of Video Based Reconstruction, this is highly dependent on combination of gaze pose and camera arrangement. This suggests changes in the VBR approach in order to allow more flexible camera arrangements. The work is of interest to those wanting to support expressive meetings that are both socially and spatially situated, and particular those using or building Immersive Virtuality Telepresence to accomplish this. It is also of relevance to the use of virtuality humans in applications ranging from the study of human interactions to gaming and the crossing of the stage line in films and TV.

No abstract

No abstract

No abstract

R e m ovi n g t h e m a s k -d o p e o pl e ov e r t r u s t a v a t a r s r e c o n s t r u c t e d fro m vid e o ?Abstract. This experiment compared the detection of deceit across video conferencing and a fixed viewpoint 3D video based computer graphic medium. The purpose was to determine if the process of 3D reconstruction influenced trust by reducing detail of facial expression. Comparison with the literature investigates the impact of facial expression on trust. Inspiration comes from previous studies in the natural and virtual world that suggest a stronger tendency to over trust a person when their facial expression is hidden. A virtual avatar that copies head and eye movement but not that of the face, could be argued as akin to a person wearing a mask. Thus, our opening research question is: Would a 3D medium that removed this mask result in a truth bias similar to video and therefore real world? Two confederates each gave a set of accounts of which half were true. These were captured and transmitted simultaneously in real time using 2D and full 3D video based communication mediums. Recordings of these sessions were later examined by two sets of participants. Twenty-one participants were asked to determine which accounts were true. Measures included: accuracy at detecting truth and deceit, and from this tendency to over trust and lastly cognitive effort in determining truthfulness. Results show that participants performed and worked to a similar degree in both mediums. Findings are of interest to those developing 3D telepresence technologies and virtual humans, and to those concerned with the trustworthiness of a medium.