Lovato Andrea, De Colle Wladimiro, Giacomelli Luciano, Piacente Alessandro, Righetto Lara, Marioni Gino, de Filippis Cosimo
Department of Neurosciences DNS, Otolaryngology Section, University of Padova, Padova, Italy.
Department of Neurosciences, Audiology Unit, Treviso Hospital, University of Padova, Treviso, Italy.
J Voice. 2016 Nov;30(6):765.e1-765.e5. doi: 10.1016/j.jvoice.2015.10.012. Epub 2016 Mar 11.
The aim of this study was to compare the discriminatory power of the Multi-Dimensional Voice Program (MDVP) and Praat in distinguishing the gender of euphonic adults.
This is a cross-sectional study.
The recordings of 100 euphonic volunteers (50 males and 50 females) producing a sustained vowel /a/ were analyzed with MDVP and Praat software.
Both computer programs identified significant differences between male and female volunteers in absolute jitter (MDVP P < 0.00001 and Praat P < 0.00001) and in shimmer in decibel (dB) (MDVP P = 0.006 and Praat P = 0.001). Using the scale proposed by Hosmer and Lemeshow, we found no gender discrimination for shimmer in dB with either the MDVP (area under the receiver operating characteristics curve [AUC] = 0.658) or Praat (AUC = 0.682). In our series, on the other hand, MDVP absolute jitter achieved an acceptable discrimination between males and females (AUC = 0.752), and Praat absolute jitter achieved an outstanding discrimination (AUC = 0.901). The discriminatory power of Praat absolute jitter was significantly higher than that of the MDVP (P = 0.003). Absolute jitter sensitivity and specificity were also higher for Praat (83% and 80%) than for the MDVP (74% and 49%).
Differences attributable to a subject's gender and to the software used to measure acoustic parameters should be carefully considered in both research and clinical settings. Further studies are needed to test the discriminatory power of different voice analysis programs when differentiating between normal and dysphonic voices.
本研究旨在比较多维度嗓音程序（MDVP）和Praat在区分嗓音悦耳的成年人的性别方面的鉴别能力。
这是一项横断面研究。
使用MDVP和Praat软件对100名嗓音悦耳的志愿者（50名男性和50名女性）发出的持续元音/a/的录音进行分析。
两个计算机程序均识别出男性和女性志愿者在绝对抖动（MDVP P＜0.00001，Praat P＜0.00001）和分贝（dB）值的闪烁方面存在显著差异（MDVP P = 0.006，Praat P = 0.001）。使用Hosmer和Lemeshow提出的量表，我们发现无论是MDVP（受试者工作特征曲线下面积[AUC] = 0.658）还是Praat（AUC = 0.682），对于dB值的闪烁均无性别鉴别能力。另一方面，在我们的研究系列中，MDVP的绝对抖动在男性和女性之间实现了可接受的鉴别（AUC = 0.752），而Praat的绝对抖动实现了出色的鉴别（AUC = 0.901）。Praat绝对抖动的鉴别能力显著高于MDVP（P = 0.003）。Praat的绝对抖动敏感性和特异性（83%和80%）也高于MDVP（74%和49%）。
在研究和临床环境中，均应仔细考虑受试者性别以及用于测量声学参数的软件所导致的差异。需要进一步研究以测试不同嗓音分析程序在区分正常嗓音和发声障碍嗓音时的鉴别能力。
