392267 Project: Designing Accessible Speech Recognition for People with Disabilities (Pj) (SoSe 2026)

Current speech recognition systems enable natural interaction with digital devices, but they often struggle to accurately process speech from people with disabilities, such as those with speech impairments or neurological conditions. This limitation reduces accessibility and prevents many users from fully benefiting from voice-based interfaces.
This project explores how speech recognition systems can be adapted to better support diverse speech patterns using modern machine learning approaches.
The goal is to design and implement a prototype system that improves recognition accuracy for atypical speech by leveraging techniques such as transfer learning, data augmentation, or personalized modeling. The system will be evaluated using appropriate speech datasets or newly collected data.

Depending on the number of students and the project scope, the project can also include:
• Evaluation of baseline speech recognition systems on atypical speech
• Development and testing of adapted or fine-tuned models
• Analysis of recognition errors and user-specific challenges
• Implementation of a simple voice interface and usability evaluation

Requirements for participation, required level

• Good programming skills in Python (ideally PyTorch)
• Basic knowledge of machine learning / deep learning
• Interest in speech processing and accessibility
• (Preferably) experience or strong interest in speech recognition or audio processing
Upon completion of this project, we will work hand in hand to publish the results in a well-established conference or journal in Human–Computer Interaction (HCI) or Computer Vision (CV)