量化和改善语音识别系统对发声障碍语音的性能
本研究基于回顾性数据库,在单一学术嗓音中心创建发声障碍语音录音数据库,测试3个语音识别平台对发声障碍语音与正常语音的性能,并训练自定义模型。结果显示所有平台在正常语音上表现良好,2个平台在发声障碍语音上显著较差;自定义模型在发声障碍语音上准确率为96.43%,正常语音为97.62%。
要点
- 所有平台在正常语音上表现良好,但2个平台在发声障碍语音上显著较差。
- IBM Watson、Amazon Transcribe和Microsoft Azure在发声障碍语音上的准确率分别为84.55%、88.57%和93.56%。
- 自定义模型在发声障碍语音上准确率为96.43%,正常语音为97.62%。
核心结论
- 当前语音识别系统对发声障碍语音的性能通常差于正常语音。
- 作者推测性能不佳是各平台原始训练数据集中缺乏发声障碍语音所致。
- 作者使用迁移学习解决这一局限,以提高这些系统在所有发声障碍语音上的性能。
研究设计与样本
- 基于回顾性数据库的实验性机器学习方法研究。
- 在单一学术嗓音中心创建发声障碍语音录音数据库。
- 将发声障碍语音输入的平台性能与正常语音输入的平台性能进行比较。
- 使用痉挛性发声障碍或声带麻痹患者的声音训练自定义语音识别模型,并将自定义模型性能与基础模型性能进行比较。
主要发现
- 所有平台在正常语音上表现良好,2个平台在发声障碍语音上表现显著较差。
- 发声障碍语音的准确率指标分别为:IBM Watson 84.55%、Amazon Transcribe 88.57%、Microsoft Azure 93.56%。
- 二次分析表明,IBM Watson和Amazon Transcribe的较低性能由痉挛性发声障碍和声带麻痹的性能驱动。
- 在Microsoft平台上构建自定义模型以提高这些病理的准确率;自定义模型在发声障碍语音上整体性能为96.43%,在正常语音上为97.62%。
关键数值与推荐
- 发声障碍语音准确率:IBM Watson 84.55%,Amazon Transcribe 88.57%,Microsoft Azure 93.56%。
- 自定义模型在发声障碍语音上准确率为96.43%,在正常语音上为97.62%。
临床怎么用
- 原文未据此给出推荐。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
OBJECTIVE: This study seeks to quantify how current speech recognition systems perform on dysphonic input and if they can be improved. STUDY DESIGN: Experimental machine learning methods based on a retrospective database. SETTING: Single academic voice center. METHODS: A database of dysphonic speech recordings was created and tested against 3 speech recognition platforms. Platform performance on dysphonic voice input was compared to platform performance on normal voice input. A custom speech recognition model was trained on voice from patients with spasmodic dysphonia or vocal cord paralysis. Custom model performance was compared to base model performance. RESULTS: All platforms performed well on normal voice, and 2 platforms performed significantly worse on dysphonic speech. Accuracy metrics on dysphonic speech returned values of 84.55%, 88.57%, and 93.56% for International Business Machines (IBM) Watson, Amazon Transcribe, and Microsoft Azure, respectively. The secondary analysis demonstrated that the lower performance of IBM Watson and Amazon Transcribe was driven by performance on spasmodic dysphonia and vocal fold paralysis. Thus, a custom model was built to increase the accuracy of these pathologies on the Microsoft platform. Overall, the performance of the custom model on dysphonic voices was 96.43% and on normal voices was 97.62%. CONCLUSION: Current speech recognition systems generally perform worse on dysphonic speech than on normal speech. We theorize that poor performance is a consequence of a lack of dysphonic voices in each platform's original training dataset. We address this limitation with transfer learning used to increase the performance of these systems on all dysphonic speech.
引用
- Hidalgo Lopez JC, Sandeep S, Wright M, et al. Quantifying and Improving the Performance of Speech Recognition Systems on Dysphonic Speech. Otolaryngol Head Neck Surg. 2023. PMID: 36939576. doi:10.1002/ohn.170
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。