Back to featured projects

A little more of the work

Computational Speech Therapy Analysis

Final-year research into speech-analysis approaches that preserve meaningful speech patterns and provide phoneme-level feedback.

Liz's Library
Back to research

Research initiative

Inside the folder

Researcher, Final-Year Project

Strathmore University · July 2021 - July 2022

Speech processingPhoneme analysisKenyan EnglishEvaluation

Description

Final-year research into speech-analysis approaches that preserve meaningful speech patterns and provide phoneme-level feedback.

Completed final-year research

The research question

What can computational speech analysis reveal when repetitions, hesitations, accent variation, and incomplete utterances are treated as information rather than noise?

My contribution

  • Developed a speech-analysis system for children with articulation difficulties, autism, and speech impairments
  • Preserved repetitions, incomplete utterances, and hesitations as potentially meaningful features
  • Evaluated phoneme error rate and F1 and investigated Kenyan English accent variation

Outcomes & evidence

  • A speech-analysis pipeline with phoneme-level feedback
  • Richer representations of speech patterns instead of automatically discarding repetitions and hesitations as noise
  • A final-year project presented at Strathmore University in 2022

Approach & methods

  • Built a speech-analysis pipeline that retained potentially meaningful speech characteristics.
  • Generated phoneme-level feedback and assessed phoneme error rate and F1.
  • Investigated how Kenyan English accent variation affected model performance.

Scope & limitations

  • The CV describes a research system, not reported clinical efficacy or a treatment outcome.
  • No diagnostic accuracy, participant count, or numerical evaluation result is added beyond the supplied evidence.

Sources & project links

Folder openResearch initiative