Associate Professor of Speech Sciences, Department of Speech-Language Sciences, All India Institute of Speech and Hearing, University of Mysore, Mysuru, Karnataka, India.
Associate Professor of Speech Sciences, Department of Speech-Language Sciences, All India Institute of Speech and Hearing, University of Mysore, Mysuru, Karnataka, India.
J Voice. 2024 Sep;38(5):1248.e1-1248.e9. doi: 10.1016/j.jvoice.2022.02.019. Epub 2022 Mar 22.
Acoustic Voice Quality Index (AVQI) has emerged in the recent past as a robust multiparametric voice quality evaluation tool. Two versions of AVQI derived using the program PRAAT have found extensive clinical and research applications. These versions have been validated in several languages around the world. However, no research reports are available on validation of AVQI in the South Indian population. Further, studies comparing the performance of the two versions of AVQI are limited in the literature.
This study was designed to validate and compare the two versions of AVQI (AVQIv02.02 and AVQIv03.01) in South Indian languages (Malayalam and Kannada).
A retrospective analysis of previously recorded voice samples was carried out on a total of 160 (91 normophonic and 69 dysphonic) voice samples. These samples were perceptually rated on a GRBAS scale by five experienced speech-language pathologists. Standardized Syllable Number (SSN) necessary to derive AVQIv03.01 was computed. Following this, these samples were analyzed to obtain the AVQIv02.02 and AVQIv03.01. The concurrent validity and diagnostic accuracy of these measures were then examined and compared.
A moderate agreement was obtained across the judges on perceptual evaluation of voice quality. SSN in Malayalam and Kannada languages were identified to be 29 and 25 syllables respectively. Language differences were not observed on both versions of AVQI. The concurrent validity of AVQIv03.01 (r = 0.788) was superior to that of AVQIv02.02 (r = 0.655). Further, the threshold of differentiating normophonic and dysphonic samples were determined to be >3.45 for AVQIv02.02 and >2.45 for AVQIv03.01.
AVQIv03.01 is superior to AVQIv02.02 in terms of its diagnostic accuracy and concurrent validity. Current findings also extend the application of AVQI as a robust tool for the evaluation of voice characteristics to the South Indian population.
声学语音质量指数（AVQI）作为一种强大的多参数语音质量评估工具，在最近出现。使用 PRAAT 程序得出的 AVQI 有两个版本，已在世界各地得到广泛的临床和研究应用。这些版本已在世界上多种语言中得到验证。然而，在印度南部人群中，尚无关于 AVQI 验证的研究报告。此外，文献中关于比较这两个版本的 AVQI 性能的研究也很有限。
本研究旨在验证和比较印度南部两种语言（马拉雅拉姆语和卡纳达语）的 AVQI 两个版本（AVQIv02.02 和 AVQIv03.01）。
对总共 160 个（91 个正常音和 69 个嗓音障碍）语音样本进行了回顾性分析。这 160 个样本由五名有经验的言语语言病理学家使用 GRBAS 量表进行感知评估。计算得出获得 AVQIv03.01 所需的标准音节数（SSN）。之后，对这些样本进行分析以获得 AVQIv02.02 和 AVQIv03.01。然后检查并比较这些指标的同时效度和诊断准确性。
在对语音质量的感知评估中，评委之间的一致性较高。在马拉雅拉姆语和卡纳达语中，SSN 分别为 29 个和 25 个音节。在两个版本的 AVQI 中均未观察到语言差异。AVQIv03.01 的同时效度（r = 0.788）优于 AVQIv02.02（r = 0.655）。此外，区分正常音和嗓音障碍样本的阈值确定为 AVQIv02.02 为>3.45，AVQIv03.01 为>2.45。
在诊断准确性和同时效度方面，AVQIv03.01 优于 AVQIv02.02。当前的发现还将 AVQI 作为评估语音特征的强大工具的应用扩展到了印度南部人群。
