Title Virtualios realybės ir dirbtinio intelekto metodai kognityvinių įgūdžių ir judesių lavinimui /
Translation of Title A study on virtual reality and artificial intelligence application to cognitive skills and movement training.
Authors Gliaubičiūtė, Donata
Full Text Download
Pages 103
Keywords [eng] rehabilitation ; virtual reality ; neural networks ; game engine ; human motion tracking
Abstract [eng] Recently, there has been a growing interest in new methods for improving the rehabilitation process, such as virtual and augmented reality. The provision of traditional rehabilitation can be expensive and difficult to reach for people living in remote regions or who do not have the opportunity to come to rehabilitation centers. Virtual reality rehabilitation systems can be accessed at patients' homes and used independently without the assistance of healthcare professionals. These systems can also be adapted to activities of daily living and improving quality of life. The application of hand motion tracking technology in virtual reality can allow users of VR systems to interact with virtual objects and situations by controlling their own hands. Data collected with motion tracking technologies can be used to recognize human activity using artificial intelligence methods. The goal of this work was to propose and investigate the application of artificial intelligence methods for the training of cognitive skills and movements using virtual reality and motion tracking technologies. The developed virtual reality rehabilitation system uses a convolutional neural network model to assign the user's movements to the appropriate activity classes. The study investigated the accuracy of the applied artificial intelligence methods, the influence of motion tracking technologies on the applied artificial intelligence methods and the usability of the developed system. The study showed that the accuracy rating of the developed model was on average 2.28% better than other existing classification models. The study found that applying Meta Quest 2 VR headset controllers can result in 2.68% higher model accuracy than using the Leap Motion Controller device. The study determined the usability of the developed system from "good" to "excellent" within the limits of the SUS scale.
Dissertation Institution Kauno technologijos universitetas.
Type Master thesis
Language Lithuanian
Publication date 2023