Project Abstract

Approximately 5–10% of the global population experience some form of speech impediment. As online, voice-based communication continues to become a vital part of everyday life—whether through virtual meetings, making friends online, or even everyday phonecalls—it is crucial that we develop inclusive tools to support individuals facing these challenges. The primary aim of my research is to create software capable of producing legible and coherent output, even when the original speech input may be partially unintelligible. This is achieved through a multi-step approach: first, by feeding the speech into a standard speech-to-text engine to capture an initial interpretation; next, by training a language model to correct or translate this output into a more accurate representation of what was likely said; and finally, by converting the corrected text back into speech using a text-to-speech system.

Keywords: Accessibility, Speech Impediment Translation, video game based implementation

 Conference Details

Session: A

Location: Sir Stanley Clarke Auditorium at 11:00 13:00

Markers: Casey Hopkins, Joe MacInnes

Course: BSc Software Engineering 3yr FT

Future Plans: I’m continuing studies