In the master thesis the process of creation and optimization of a 3D virtual interpreter of Slovenian Sign Language is presented, intended for semi-automated interpretation for a limited vocabulary with usage of Application Programming Interfaces. The Blender environment enables the creation, rigging and animation of advanced photorealistic 3D models, as well as the management of pre-prepared actions using Python API. Due to the goal to bring final renders as close as possible to real time rendering, several optimizations were performed, from optimization of the polygon mesh, the created rig for more efficient movement, created API adjustments, as well as an optimisation of the render engine settings and analysis of render speed on different hardware equipment.
The use of a 3D virtual interpreter can be a useful alternative to increase accessibility, especially in emergencies, such as the outbreak of the COVID-19 pandemic. Due to the relatively high demands of the deaf community which requires virtual interpreters to move as naturally as possible, with smooth gestures and without unnecessary returns to a neutral position, and to show eloquent facial expressions, the creation of an appropriate solution requires in-depth analysis of all key areas of the master thesis. The final renders in the practical part proved to be successful, as all users in the preliminary study easily understood all animated signed sentences. The final creation could be used in several different domains, while it could also be utilised, in collaboration with the deaf community, as a semi-automatic translator of spoken Slovenian to sign language.
|
