Acoustic Measurement App
- PC based measurement not being utilised in the clinic
- Measurement algorithms and technique reproduced in a mobile app
- Patients/clients can run the app in the waiting room, so data collection is activated in this setting
Data Science & Research Support
Projects
From the desktop to the waiting room: making data collection mobile.
Optimizing the audio pipeline: replacing general-purpose editing with a specialised single-click segmentation and automated naming tool.
Audio Processing Workflow Improvements
- Audio waveforms required to be parsed, measured and edited
- Built GUI tool for timestamping & annotation
- Workflow speed and accuracy greatly improved
Ensuring data privacy: local deployment of speech-to-text and speaker identification for sensitive clinical interviews.
Locally executed Automatic Speech Transcription and Speaker Identification
- The dataset comprised videos of patient interviews
- These were sensitive and online transcription was to be avoided
- I built a local-machine solution using OpenAI's Whisper model, incorporating Automatic Speech Recognition (ASR), timestamping and diarisation by identifying multiple speakers
