发表机构
National Electronics and Computer Technology Center (NECTEC); King Mongkut’s University of Technology Thonburi; King Mongkut’s Institute of Technology Ladkrabang（国家电子与计算机技术中心; 国王蒙固科技大学吞武里分校; 国王蒙固科技学院叻甲邦分校）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究构建28小时缅甸医疗语音语料库，通过全微调（FFT）和带LoRA的PEFT微调Whisper，提升其在医疗对话ASR中的性能，最优模型WER达23.44%，优于更大的通用领域微调模型。
AI 中文摘要
尽管Whisper模型受益于大规模多语言预训练，但其在缅甸医疗语音上的表现仍有限。本研究提出一种缅甸医疗语音识别框架，基于由母语者录制并验证的高质量28小时语料库构建。我们使用全微调（FFT）和带LoRA的参数高效微调（PEFT）对Whisper模型进行微调。为评估鲁棒性，我们在受控噪声和模拟房间声学条件下应用波形级和频谱图级数据增强。尽管增强会降低纯净语音上的性能，但在FFT和PEFT设置下，其显著提升了模型在噪声和混响环境中的鲁棒性。我们表现最佳的系统为未使用增强的全微调myMediWhisper-Medium，达到了23.44%的先进词错误率（WER），优于大得多的通用领域微调模型。数据集及其他资源可在Huggingface仓库获取：this https URL。
英文摘要
Although Whisper models benefit from large-scale multilingual pre-training, their performance on Burmese medical speech remains limited. This work presents a Burmese medical speech recognition framework built on a high-quality 28-hour corpus recorded and validated by native speakers. We fine-tune Whisper models using full fine-tuning (FFT) and parameter-efficient fine-tuning (PEFT) with LoRA. To evaluate robustness, we apply waveform- and spectrogram-level data augmentation under controlled noise and simulated room acoustics. While augmentation reduces performance on clean speech, it significantly improves robustness in noisy and reverberant environments across FFT and PEFT settings. Our best-performing system, fully fine-tuned myMediWhisper-Medium without augmentation, achieves a state-of-the-art Word Error Rate (WER) of 23.44%, outperforming much larger general-domain fine-tuned models. Dataset and other resources can be found at the Huggingface repository: https://huggingface.co/datasets/LULab/mediTalk-mm-rdy.
