Impact of different type of Child Avatar Interactions on user Quality of Experience
Abstract
Conducting an interview and communicating with children that have experienced traumatic situations can be difficult. Norway’s Child Protective services received in 2017 over 58.580 reports about child maltreatment and estimated that over 118 million children in Europe are victim to abuse.
Gunn Astrid Baugerud and her team are creating an avatar in virtual reality that will work as a training avatar for police to work on their interviewing abilities. This thesis will improve this avatar by testing multiple techniques and technologies that can be applied to the avatar. The thesis will evaluate two different solutions made with Unity and one artificially generated avatar with generative adversarial network techniques. The thesis is however limited to only Unity based avatars and movement testing of movement have been restricted due to corona-19 restrictions that was applied during the thesis.
The testing of the three different solution gave indications that avatars created with generative adversarial network techniques had the best impact on realism and overall experience. The avatars created with Ready Player Me follow closely, with minor differences when it comes to appearance and experience. The worst avatar was made with Unity multipurpose avatar 2, during the questionnaire one of the avatars created with UMA2 had the worst results of all avatars.
The avatar created in the main project by Gunn Astrid Baugerud and her team has come a great way, the results of this thesis show what could be the focus points when creating the final virtual child avatar, with post-questionnaire indicating that eye contact and graphics are among the most important attributes for accomplishing realism.