Abstract
The vision of this master’s thesis is to reduce the communication gap between hearingand deaf individuals. To achieve this, the technical capabilities in the fields of machine
learning and neural networks were explored and implemented in a prototype for sign
language recognition. The communication barrier between hearing and deaf people leads
to a significant separation. Technology should always support people and in this case it
should facilitate communication. This thesis also aims to spark interest in sign language.
A mobile app, seamlessly integrated with a smartphone camera and responsive to gesture interaction, could captivate individuals who have not previously engaged with sign
language. Users can enter this field playfully through gamification as they have to learn
the corresponding gesture first before they can interact with the app.
In an era where the Turing Test has become a reality, we can communicate with
ChatGPT in a natural, human-like manner, such technologies should undoubtedly be
employed to promote new ways of interpersonal communication. The thesis highlights
the critical importance of training data for machine learning. Specifically, substantial
efforts are still needed in the collection of gesture videos to enable smooth humancomputer interaction through sign language. Developments in machine learning are
advancing rapidly. Even during the preparation of this thesis, a study was published
proposing the Kolmogorov-Arnold representation theorem for machine learning. This
thesis should also raise the awareness that machine learning models are dynamic constructs. While a perfect model will never exist, the approximation to it can continually
improve. The chapter “Implementation and discussion” and the experiments therein
document the initial steps in such an iterative development process.
| Date of Award | 2024 |
|---|---|
| Original language | German (Austria) |
| Supervisor | Werner Christian Kurschl (Supervisor) |