Get Started
Home
Topics
Search
Library
Research questionWhich PEFT variant and enrollment budget best balance accuracy, storage, and training practicality for per-patient dysarthric ASR?Dysarthric ASR must adapt to an individual speaker, but available enrollment speech may be limited and storing a separate adapter for each patient can be costly. The choice of PEFT variant affects both recognition gains and the practical cost of personalization.
AI
Audio & Speech
Audio & Speech Processing
Evaluation & Benchmarks
Health
Machine Learning
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.Choosing a PEFT Variant for Per-Patient Dysarthric ASR: A Single-Speaker Case Study on Two ASR BasesEvidence comes from a single post-stroke Hungarian male speaker with severe dysarthria and 409 utterances. The study compares LoRA, QLoRA, AdaLoRA, DoRA, LoHA, VeRA, and VB-LoRA on Whisper-large-v3 with Hungarian fine-tuning and a multilingual Qwen3-ASR-1.7B checkpoint, using CER, adapter size, and training considerations; enrollment effects are measured across 0–30 minutes of patient audio. Results are limited to this speaker and language, with no evidence for broader populations.research paper · Sep 2, 2026
Related questions
How can speech enhancement adapt to mismatched deployment acoustics without labeled target audio?How can EEG decoding remain reliable across subjects and sessions for assistive target selection?How can spoken language detection adapt to underrepresented accents under low-resource constraints without overfitting?How can speaker-attributed ASR identify who said what as speech arrives with low latency?