Continual Learning for User Identification and Game Classification Based on Virtual Reality Headset Data: Applications on the Questset Dataset

User identification is a fundamental component for enabling personalization and adaptive interactions in Virtual Reality (VR) systems. This thesis investigates the applicability of Continual Learning (CL) methods to user identification. The study uses motion data from the publicly available Questset dataset, which is constructed from data collected with the Meta Quest 2 VR headset. The motivation for applying CL lies in its ability to allow models to adapt to new users and behavioral patterns, without discarding previously acquired knowledge. First, this work evaluates the effectiveness of recurrent neural networks (RNNs) for user identification within a standard Deep Learning framework. Then, building on the obtained results, the study is extended to a CL setting to assess the feasibility and benefits of incremental adaptation for the VR user identification problem.