A fine-tuned Whisper tiny.en model reaches 15.9% WER on children's speech and runs in real time on a Raspberry Pi, with low-rank compression trading an 11% relative WER rise for lighter computation.
Model Training and Evaluation (on GPU) All training (fine-tuning) and evaluation recipes were adapted from HuggingFace and OpenAI
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Adapting Whisper for Lightweight and Efficient Automatic Speech Recognition of Children for On-device Edge Applications
A fine-tuned Whisper tiny.en model reaches 15.9% WER on children's speech and runs in real time on a Raspberry Pi, with low-rank compression trading an 11% relative WER rise for lighter computation.