Periodic Reporting for period 1 - ForEVR (Appealing Character Design for Embodied Virtual Reality)
Período documentado: 2021-09-01 hasta 2023-08-31
The overall objective of the ForEVR project was to create appealing characters which will be appropriate for serious, non-entertainment applications in virtual reality (VR). In order to create appeal, most designers resort to using stereotypical representations of characters which can be harmful to specific populations of users. This can be especially problematic when using the characters for non-serious applications, such as virtual therapy with vulnerable subjects. In order to resolve this issue, the ForEVR project investigated the ability of the character body motion to convey appealing personalities and non-threatening behaviour. The project aimed to create an embodied VR (EVR) protocol to produce a highly immersive and transformative experience, able to induce strong emotional and behavioural response to the character and situation. The project has achieved this overall objective by identifying particular features in human motion which affect human behaviour in EVR which indicate higher/lower levels of appeal of the virtual character. The results of this project specifically aid the development of better scenarios for VR therapy and training but are also relevant for the emerging communication platforms, such as the Metaverse.
In the duration of the ForEVR project, three main objectives were addressed. First was the creation of EVR protocols and environments for the purposes of perceptual experimentation. Additionally, an existing EVR environment and protocol focusing on mental health topics, named the ConVRself, created by the Event Lab (University of Barcelona) and the secondment partner (Virtual Bodyworks) was considered.
The second objective was to explore the most appropriate techniques for manipulating recorded human biological motion. We focused on the OCEAN model personality traits (neuroticism, emotional stability), which can be identified from body and facial motion of people and could be associated with higher or lower levels of appeal. We then investigated three different approaches to creating the motion filter, which will change the existing motion towards a more appealing/less appealing style for virtual characters (Laban motion analysis descriptors, entropy in motion, and style transfer techniques using machine learning). The preliminary study, where we investigated the relationship between appearance and motion of virtual characters was also conducted as part of the second objective. We found that motion as well as appearance both separately contribute to the appeal of the character, giving an indication that appearance needs to be controlled when investigating the motion perception in our final experiment.
Lastly, we evaluated how the final character design affected the users’ perception and propose character design guidelines for improving the EVR experience. We created two characters, who exhibited different personality types through their motion movement patterns and displayed them in an immersive experiment where we instructed the participants to complete a choice task which included avoiding the two characters in a tight space (realistic metro scene). We were able to receive evidence of perceptual and behavioural changes of users in interaction with virtual agents with different movement patterns. The ‘neurotic’ motion pattern resulted in a larger distance and avoidance of the character, as well as choosing alternative exit doors when the neurotic type was obstructing one of the exits. We published our findings in the journal Computers & Graphics and IEEEVR 2022 proceedings.
The second objective was to explore the most appropriate techniques for manipulating recorded human biological motion. We focused on the OCEAN model personality traits (neuroticism, emotional stability), which can be identified from body and facial motion of people and could be associated with higher or lower levels of appeal. We then investigated three different approaches to creating the motion filter, which will change the existing motion towards a more appealing/less appealing style for virtual characters (Laban motion analysis descriptors, entropy in motion, and style transfer techniques using machine learning). The preliminary study, where we investigated the relationship between appearance and motion of virtual characters was also conducted as part of the second objective. We found that motion as well as appearance both separately contribute to the appeal of the character, giving an indication that appearance needs to be controlled when investigating the motion perception in our final experiment.
Lastly, we evaluated how the final character design affected the users’ perception and propose character design guidelines for improving the EVR experience. We created two characters, who exhibited different personality types through their motion movement patterns and displayed them in an immersive experiment where we instructed the participants to complete a choice task which included avoiding the two characters in a tight space (realistic metro scene). We were able to receive evidence of perceptual and behavioural changes of users in interaction with virtual agents with different movement patterns. The ‘neurotic’ motion pattern resulted in a larger distance and avoidance of the character, as well as choosing alternative exit doors when the neurotic type was obstructing one of the exits. We published our findings in the journal Computers & Graphics and IEEEVR 2022 proceedings.
Apart from the main results, the research in the project also explores new methods to evaluate the perception of virtual characters by studying the behaviour of the users in VR (analysis of the walking trajectory, choice task), which can be used by other researchers interested in studying the perception of virtual characters. The guidelines which can be created on the basis of the results of this project have an impact on the society. They can be used to create new rehabilitation protocols which manipulate the motion of the virtual characters to influence people’s behaviour and aid (or challenge) the therapeutical outcome. The research conducted in this project provides guidelines on how to create simple variation in behaviour for companies which use virtual humans for their training/rehabilitation. The knowledge on the aspects of motion which aid the perception of appeal in virtual characters can be applied to the mental health protocols of EVR and have the potential to increase the positive outcomes of the treatment. With the visibility of this project, the importance on developing tools for mental health treatment using EVR is emphasised, bringing this possibility to the general public, who may not be aware that such tools exist.