Saved in:
Bibliographic Details
Main Authors: Bao, Chen, Huo, Chuanbing, Chen, Qinyu, Gao, Chang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.06566
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911415278764032
author Bao, Chen
Huo, Chuanbing
Chen, Qinyu
Gao, Chang
author_facet Bao, Chen
Huo, Chuanbing
Chen, Qinyu
Gao, Chang
contents This paper proposes AS-ASR, a lightweight aphasia-specific speech recognition framework based on Whisper-tiny, tailored for low-resource deployment on edge devices. Our approach introduces a hybrid training strategy that systematically combines standard and aphasic speech at varying ratios, enabling robust generalization, and a GPT-4-based reference enhancement method that refines noisy aphasic transcripts, improving supervision quality. We conduct extensive experiments across multiple data mixing configurations and evaluation settings. Results show that our fine-tuned model significantly outperforms the zero-shot baseline, reducing WER on aphasic speech by over 30% while preserving performance on standard speech. The proposed framework offers a scalable, efficient solution for real-world disordered speech recognition.
format Preprint
id arxiv_https___arxiv_org_abs_2506_06566
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AS-ASR: A Lightweight Framework for Aphasia-Specific Automatic Speech Recognition
Bao, Chen
Huo, Chuanbing
Chen, Qinyu
Gao, Chang
Audio and Speech Processing
Artificial Intelligence
This paper proposes AS-ASR, a lightweight aphasia-specific speech recognition framework based on Whisper-tiny, tailored for low-resource deployment on edge devices. Our approach introduces a hybrid training strategy that systematically combines standard and aphasic speech at varying ratios, enabling robust generalization, and a GPT-4-based reference enhancement method that refines noisy aphasic transcripts, improving supervision quality. We conduct extensive experiments across multiple data mixing configurations and evaluation settings. Results show that our fine-tuned model significantly outperforms the zero-shot baseline, reducing WER on aphasic speech by over 30% while preserving performance on standard speech. The proposed framework offers a scalable, efficient solution for real-world disordered speech recognition.
title AS-ASR: A Lightweight Framework for Aphasia-Specific Automatic Speech Recognition
topic Audio and Speech Processing
Artificial Intelligence
url https://arxiv.org/abs/2506.06566