Project Kasa

This project aims to develop an automatic speech recognition (ASR) model tailored for non-standard African speech, with a focus on individuals with speech impairments such as dysarthria, stroke and cerebral palsy.

Existing ASR systems perform poorly on African-accented English, code-switched speech, and atypical speech patterns due to limited data and high acoustic variability.

Leveraging transfer learning and personalized adaptation techniques, the project adapts state-of-the-art models to accurately recognize accented, code-switched, and impaired speech. The ultimate goal is to enable inclusive, reliable, and privacy-preserving voice-based assistive technologies for African users.