Joint fine-tuning on seven dysarthric speakers' data reduced per-speaker character error rates by up to 13.15 percentage points compared to single-speaker fine-tuning on the CDSD corpus.
Furtheranalysisparadoxically showed that sequential fine-tuning—first with multi-speaker data followed by speaker-specific refinement—increased CERs for Speakers 04 and 06
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Cross-Learning Fine-Tuning Strategy for Dysarthric Speech Recognition Via CDSD database
Joint fine-tuning on seven dysarthric speakers' data reduced per-speaker character error rates by up to 13.15 percentage points compared to single-speaker fine-tuning on the CDSD corpus.