Parameter-based and embedding-based averaging in personalized FL for dysarthric ASR yields up to 0.99% absolute WER reduction on UASpeech and 0.56% on TORGO versus regularized FedAvg.
Phone-purity Guided Discrete Tokens for Dysarthric Speech Recognition,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Towards Personalized Federated Learning for Dysarthric Speech Recognition
Parameter-based and embedding-based averaging in personalized FL for dysarthric ASR yields up to 0.99% absolute WER reduction on UASpeech and 0.56% on TORGO versus regularized FedAvg.