This thesis explores challenges for the design of Virtual Reality Exposure Therapy (VRET) systems. Exposure therapy is the established method for treatment of anxiety disorders and is typically delivered in-vivo, i.e. exposure to phobic stimulus in real environments. Virtual reality (VR), instead, offers the potential to conduct exposure therapy at the clinic. This approach has several benefits in terms of efficiency, customization and control, amount of exposure, and as an transition phase to real situations. However, currently many systems are limited in scope and are designed for research purposes without informing the design from therapist's practices.  My research aims to contribute towards the understanding of current practices in exposure therapy and investigates challenges for the design of these systems for the two main user groups, patients and therapists. Three different focus areas have been prevalent. First, we have studied therapist in real sessions to inform the design and development of VRET-systems. Second, we have evaluated two different VRET implementations supporting therapists to interact with patients. Third, on the patient's side, we have studied presence on healthy participants focusing on the influence of virtual bodies and patient movement in VR. This thesis summarises and discusses these studies. Overall, the studies emphasize the complexity of exposure therapy and the need for individualized patient conditions. This poses multiple challenges for the design of VRET-systems such as, first, the systems must offer flexibility to the therapists to orchestrate individualized therapy. Second, the systems must enable rich therapists-patient interaction. Third, the complexity of individualization of scenarios and sessions must be addressed in the design of the therapist's interface. Fourth, for patients, body avatars influences presence differently depending on the scenario and locomotion is challenging as offices are typically small. / <p>QC 20190214</p>

No description available.

No description available.

No description available.

The previous decade has seen breakthroughs in image analysis and computer vision, mainly due to machine learning methods known as deep learning. These methods have since spread to other fields. This thesis aims to survey the progress, highlight problems related to data and computations, and show techniques to mitigate them. In Paper I, we show how to modify the VGG16 classifier architecture to be equivariant to transformations in the p4 group, consisting of translations and specific rotations. We conduct experiments to investigate if baseline architectures, using data augmentation, can be replaced with these rotation-equivariant networks. We train and test on the Oral cancer dataset, used to automate cancer diagnostics. In Paper III, we use a similar methodology as in Paper I to modify the U-net architecture combined with a discriminative loss, for semantic instance segmentation. We test the method on the BBBC038 dataset consisting of highly varied images of cell nuclei. In Paper II, we look at the UCluster method, used to group sub- atomic particles in particle physics. We show how to distribute the training over multiple GPUs using distributed deep learning in a cloud environment. The papers show how to use limited training data more efficiently, using group-equivariant convolutions, to reduce the prob- lems of overfitting. They also demonstrate how to distribute training over multiple nodes in computational centers, which is needed to handle growing data sizes.

As an increasingly powerful, interactive, and dynamic medium for sharing information, the World Wide Web (Web) has found many educational applications. Because of educational benefits for both learners and educators, Web-based courses and programs have increasingly been developed and implemented by many academic institutions, organizations, and companies worldwide (Aggarwal & Brento, 2000). This study, however, asserts that many of the developmental approaches lack three important considerations to be used for implementing learning applications based on the Web: (1) integration of the human-computer interface design with instructional design, (2) development of the evaluation framework to improve the overall quality of Web-based learning support environments, and (3) development of an individualistic approach to the Web hosting of courses. This study explored the three Web-based learning environment weaknesses while developing a user-centered, Web-based learning support environment for Global Positioning System (GPS) education: Web-based distance and distributed learning (WD2L) environment. Research goals of the study were all concerned with the improvement of the design process and usability of the WD2L environment based on a theory-based Integrated Design Process (IDP) proposed in the study. Results indicated that the proposed IDP was effective in that the study showed (1) the WD2L environment's equivalence to traditional supplemental learning, especially as a Web-based supplemental learning program and (2) users' positive perceptions of WD2L environment resources. The study also confirmed that for an e-learning environment to be successful, various aspects of the learning environment as a Web-based supplemental learning program should be considered such as application domain knowledge (i.e., target subject field), conceptual learning theory, instructional design, human-computer interface design, and evaluation about the overall quality of the learning environment. / Ph. D.

Responding to the need within the human-computer interaction field to address ubiquitous and multitasking systems more scientifically, this research extends the usefulness of a new research framework for a particular class of systems. Notification systems are interfaces used in a divided-attention, multitasking situation, attempting to deliver current, valued information through a variety of platforms and modes in an efficient and effective manner. Through review of literature and experiences with empirical dual-task perceptual studies, we recognize a lack of unifying framework for understanding, classifying, analyzing, developing, evaluating, and discussing notification systems--fundamentally inhibiting scientific growth and knowledge reuse that should help designers advance the state-of-the-art. To this end, we developed a framework (referred to as the IRC framework) for notification systems research based on a core taxonomy of critical parameters describing user goals. Next, we extend the framework, focusing on three key aspects: 1) a system description process, allowing articulation of abstract design objectives that focus on critical user requirements; 2) interface usability evaluation tools, enabling comparison of the design and user's models, while supporting generalizability of research and early identification of usability concerns; and 3) design comparison and reuse mechanisms, saving time and effort in requirements analysis and early design stages by enabling design reuse and appreciation of design progress. Results from this research include the development of tools to express IRC design models (IRCspec) and user's models (IRCresults), and the extension of the critical parameters concept. Validation studies with novice designers show sufficient assessment accuracy and consistency. Leveraging these tools that help designers express abstract, yet critical, design intentions and effects as classification and retrieval indices, we develop a repository for reusable design knowledge (a claims library). Responding to challenges of design knowledge access that we observed through initial user testing, we introduce a vision for an integrated design environment (LINK-UP) to operationalize the IRC framework and notification systems claims library in a computer-aided design support system. Proof-of-concept testing results encourage the thought that when valuable design tools embody critical parameters and are coupled with readily accessible reusable design knowledge, interface development will improve as a scientific endeavor. / Ph. D.

This dissertation focuses on taking a human-centric approach to utilize human intelligence best to inform machine learning models. More specifically, the complex relationship between the changes in movement functionality to movement quality. I designed and evaluated the Tacit Computable Empowering methodology across two domains: in-home rehabilitation and clinical assessment. My methodology has three main objectives: first, to transform tacit expert knowledge into explicit knowledge. Second, to transform explicit knowledge into a computable framework that machine learning can understand and replicate. Third, synergize human intelligence with computational machine learning to empower, not replace, the human. Finally, my methodology uses assistive interfaces to allow clinicians and machine learning models to draw parallels between movement functionality and movement quality. The results from my dissertation inform researchers and clinicians on how best to create a standardized framework to capture and assess human movement data for embodied learning scenarios / Doctor of Philosophy / Artificial intelligence (AI) is increasingly considered an important computational design material in the development of innovative products, systems, and services. Recent research emphasizes the potential for computational designers to create new tools, methods, and design processes to more adeptly handle AI and machine learning as fundamental but not exclusive materials within the design process. This talk adopts a human-centric approach to utilize human intelligence to inform machine learning models within a healthcare context. I describe the novel tacit computable empowering (TCE) methodology used and evaluated across two healthcare domains: in-home rehabilitation and clinic-based assessment. The TCE methodology comprises three main objectives: 1) to transform tacit expert knowledge into explicit knowledge; 2) to transform explicit knowledge into a computable framework that machine learning can understand and replicate and 3) to synergize human intelligence with computational machine learning to empower (and not replace) the human. This methodology uses assistive interfaces to allow clinicians and machine learning models to draw parallels between movement functionality and movement quality. Outcomes from this work inform researchers and clinicians as to how to best create a standardized framework to capture and assess human movement data for embodied learning scenarios.

Background: Physical inactivity of the general population is a major public health concern in the US and around the world. Community-based interventions that include evidence-based principles of group dynamics are effective at improving individual-level physical activity behaviors as well as changing social norms for health behaviors. The use of technologies such as smartwatches has a potential to channel and amplify underlying group dynamics principles in such interventions. In order to explore the use of smartwatches for group dynamics-based physical activity interventions, a smartwatch centered system was designed and deployed as part of an eight-week pilot study. Objectives: The primary goal of this study was to explore the degree to which smartwatches effectively channel group dynamics strategies in the context of an eight-week community based physical activity intervention. Methods: In this explanatory mixed-methods study, system usage data were analyzed (e.g., frequency of interaction with smartwatch and smartphone) and participant physical activity (e.g., participant steps tracked by the system). To provide a richer picture of the user experience, use of features, and impact of group dynamics, participants were invited to participate in one-on-one interviews after the pilot program. The group dynamics-based questions centered on the individual’s attraction to the group task and socially as well as the individual’s perception of group integration around the task and as a social unit (i.e., the four dimensions of cohesion). The interview recordings were transcribed verbatim and analyzed via an abbreviated grounded theory approach. The system usage data was visually and numerically summarized. Results: Five of the seven participants completed interviews. The interview analysis resulted in 365 meaning units representing 2 themes (related to user experience with devices and manifestations of group dynamic principles), 4 sub-themes and 23 categories. The participants completed 31.3 (SD=2.91) miles per week and engaged with the smartwatch and the Android app 2.6 and 1.5 times a day, respectively. Analysis of interviews and the system usage logs from five participants, reveal sustained engagement with the smartwatch and the smartphone app. The system facilitated self-reflection and awareness in terms of physical activity levels, encouraged interactions with the team members and helped them to stay aware of the daily goals. Additionally, the participants reported habit formation in terms of wearing and using the smartwatch on the daily basis. Implications: This study provides preliminary support that accessible information via the custom smartwatch watchface can be a viable solution for retaining higher participant engagement during group dynamics-based community interventions. Such devices can help expand group-dynamics interventions by making them less depended on in-person delivery methods. / Master of Science / Physical inactivity of the general population is a major public health concern in the US and around the world. Community-based interventions, with group dynamics strategies at the core, are effective at improving individual physical activity behaviors. The use of technologies such as smartwatches has potential to channel and amplify the underlying program principles in such interventions. This work presents a smartwatch-centered system to encourage group cohesion in physical activity interventions, exploring it as part of an eight-week study that revealed participant awareness of group performance through smartwatch interactions.

The research presented in this thesis concerns the contribution of virtual human (or avatar) fidelity to social interaction in virtual environments (VEs) and how sensory fusion can improve these interactions. VEs present new possibilities for mediated communication by placing people in a shared 3D context. However, there are technical constraints in creating photo realistic and behaviorally realistic avatars capable of mimicking a person's actions or intentions in real time. At the same time, previous research findings indicate that virtual humans can elicit social responses even with minimal cues, suggesting that full realism may not be essential for effective social interaction. This research explores the impact of avatar behavioral realism on people's experience of interacting with virtual humans by varying the interaction fidelity. This is accomplished through the creation of Cinemacraft, a technology-mediated immersive platform for collaborative human-computer interaction in a virtual 3D world and the incorporation of sensory fusion to improve the fidelity of interactions and realtime collaboration. It investigates interaction techniques within the context of a multiplayer sandbox voxel game engine and proposes how interaction qualities of the shared virtual 3D space can be used to further involve a user as well as simultaneously offer a stimulating experience. The primary hypothesis of the study is that embodied interactions result in a higher degree of presence and co-presence, and that sensory fusion can improve the quality of presence and co-presence. The argument is developed through research justification, followed by a user-study to demonstrate the qualitative results and quantitative metrics.This research comprises of an experiment involving 24 participants. Experiment tasks focus on distinct but interrelated questions as higher levels of interaction fidelity are introduced.The outcome of this research is the generation of an interactive and accessible sensory fusion platform capable of delivering compelling live collaborative performances and empathetic musical storytelling that uses low fidelity avatars to successfully sidestep the 'uncanny valley'. This research contributes to the field of immersive collaborative interaction by making transparent the methodology, instruments and code. Further, it is presented in non-technical terminology making it accessible for developers aspiring to use interactive 3D media to pro-mote further experimentation and conceptual discussions, as well as team members with less technological expertise. / Master of Science