| Abstract [eng] |
This dissertation investigates virtual reality (VR) control methods for peripheral device integration and human posture analysis using off-the-shelf VR controls. The research addresses challenges arising from the use of control inputs from devices not originally designed for VR, including data transmission latency, synchronisation inaccuracies, and visual stuttering, which affect the handling and interpretation of input data in VR systems. Data prediction methods such as interpolation, extrapolation, and filtering techniques are analysed and experimentally evaluated using an experimental virtual rowing system to assess their suitability for controlling motion in VR under controlled conditions. In addition, the dissertation examines control methods based on off-the-shelf VR tracking for human posture monitoring and movement analysis by presenting a processing workflow that includes positional data acquisition, transformation into joint-angle representations, and the application of machine learning methods for exercise detection and movement correctness classification. Random Forest and Convolutional Neural Network models are evaluated using eight predefined upper-body exercises, and the results provide an assessment of the applicability of VR-based control and data processing methods for posture evaluation in controlled experimental environments. |
