Dinithi Dissanayake

We conducted a systematic literature review of 63 papers examining how adaptive wearables sense user states, trigger context-aware interventions, and support real-time cognitive or behavioral feedback. Our review introduces a taxonomy of sensing modalities, adaptation triggers, and intervention strategies, and highlights key design challenges and opportunities. This work provides a foundation for developing next-generation wearables that meaningfully adapt to users’ needs. This paper is currently under review for publication.

Sensory Spotlight explores how AI can anticipate shifts in human attention by combining audio and visual signals—much like how we naturally react to sudden events in our environment. In the example above, a loud noise (a phone dropping) is detected through audio, and the system predicts that the user’s attention will shift toward the floor. Our predictive model not only predict user attention but also provides saliency to determine which modality got the spotlight. This helps to take adaptive decisions--where information should appear inside a smart-glasses display and whether feedback should be delivered visually or through audio cues. By dynamically selecting the most appropriate output modality, Sensory Spotlight brings us closer to intelligent, context-aware wearables that adapt to how we perceive the world.

VRsense is an explainable system developed to help VR game developers assess cybersickness in their games. Unlike traditional black-box approaches, our system utilizes engineered features to provide actionable insights into game design and user interactions. Designed to be plug-and-play, VRSense allows any VR game developer to upload their data or run VRHook on their gameplay, receiving valuable feedback on how effectively their game tackles motion sickness. Read our paper for more details.

We developed a novel, state-of-the-art 3D-2D correspondence technique to enhance the understanding of 3D point clouds. Additionally, we created a 3D vision algorithm that allows users to add, delete, or modify parts of 3D objects, enabling the regeneration of transformed objects on the other end. Our solution was rigorously evaluated against a series of simulated privacy attacks. To assess its real-world applicability, we implemented the algorithm on a smart device, testing its practicality for deployment in Mixed Reality environments. This is the Final Year Project of my Undergraduate Studies. This project resulted in two papers: 3D representation learning from 3D:2D and vision algorithms.

The study explored how generative AI models, such as ChatGPT, contribute to value co-creation and examined how the attitudes and impressions of non-technical users influence their adoption of such technologies. This re- search aimed to understand how non-technical users interact with and respond to the current generative AI hype in the domain of primary school education. Interviewed teachers and developers centered around the adoption of a genAI-based education tool named Yuni/Hopu.