Augmented Reality for Cognitive Impairment and Dementia

Abstract

Dementia is part of the normal ageing process and there are very few Augmented Reality (AR) technologies to help people with dementia. The implementation of these technologies is not fully successful due to limitations in the accuracy of image processing and interactive instruction. This paper aims to improve the accuracy and interactivity of the system to perform tasks in context-specific scenarios. Our objective is to increase the accuracy of video and the interactive instruction of the system by minimizing blurry image processing and overlaying the 3D images over 2D objects respectively. The recommended system involves Optional Frame Selection Algorithm (OFSA), with an 8D descriptor vector in speed up robust features (ESURF) to increase the accuracy of image processing, and overlay interactive 3D images to the original image to make the augmented instruction more effective. The result shows that the purposed (ESURF) algorithm utilizes more details and features in the descriptor vector to improve the regular speed up robust features (SURF) algorithm that comes from filtered OFSA. This result of performance shows that the ESURF algorithm outperforms regular SURF detectors in accuracy for matching objects. A usability study suggested that augmented interactive 3D images are more effective than just providing text as instruction. The proposed system emphasises on reducing blurry images for processing, maximising matching descriptors to increase accuracy during image processing, and to increase the interactivity instruction of the system, which provides dementia patients with more of a sense of the real environment and more effective instruction.