Departamento de Fonoaudiologia, Faculdade de Medicina, Universidade Federal de Minas Gerais - UFMG - Belo Horizonte (MG), Brasil.
Departamento de Engenharia Eletrônica, Escola de Engenharia, Universidade Federal de Minas Gerais - UFMG - Belo Horizonte (MG), Brasil.
Codas. 2021 May 3;33(1):e20190197. doi: 10.1590/2317-1782/20202019197. eCollection 2021.
To analyze if the use of synthesized voice anchor emissions in auditory-perceptual evaluation improves intra- and inter-rater agreement.
This is a quantitative study. Thirty-two inexperienced evaluators were selected and performed two activities on a Programming Interface created by the authors: Active Calibrator Activity - auditory-perceptual evaluation of the roughness and breathiness parameters as 0-no deviation, 1-slight deviation, 2-moderate deviation, or 3-intense deviation of 25 voices with the support of anchored emissions of synthesized voices; and Inactive Calibrator Activity - auditory-perceptual evaluation of these same voices without the support of anchored vocal emissions. The voices were randomized for each activity, and the order of the activities was drawn randomly for each evaluator. The second activity was performed 15 days after the first. The Kappa coefficient was used to analyze intra- and inter-rater agreement, and the confidence interval (CI) was employed to compare concordances.
Inter-rater agreement was higher for the intense degree of the breathiness parameter in the Active Calibrator Activity when compared to the Inactive Calibrator Activity, as well as the intra-rater agreement of the roughness parameter.333.
Use of anchor emissions of synthesized voices directly in the evaluation improves intra- and inter-rater agreement in auditory-perceptual voice analysis.
分析在听觉感知评估中使用合成语音锚定发声是否能提高内部和外部评估者之间的一致性。
这是一项定量研究。选择了 32 名没有经验的评估者，并在作者创建的编程接口上进行了两项活动：主动校准器活动——对粗糙度和呼吸音参数进行听觉感知评估，25 个声音的锚定发声支持下，将偏差程度评为 0-无偏差、1-轻微偏差、2-中度偏差或 3-重度偏差；以及非主动校准器活动——在没有锚定发声支持的情况下对这些相同的声音进行听觉感知评估。每个活动中的声音都是随机的，每个评估者的活动顺序也是随机抽取的。第二项活动在第一项活动后 15 天进行。使用 Kappa 系数分析内部和外部评估者之间的一致性，并使用置信区间（CI）比较一致性。
在主动校准器活动中，呼吸音参数的重度偏差的外部评估者之间的一致性高于非主动校准器活动，以及粗糙度参数的内部评估者之间的一致性为 0.333。
在评估中直接使用合成语音的锚定发声可以提高听觉感知语音分析中的内部和外部评估者之间的一致性。
