The Acoustic Voice Quality Index Version 03.01 for the Japanese-speaking Population

Publication date: Available online 15 November 2017
Source:Journal of Voice
Author(s): Kiyohito Hosokawa, Ben Barsties v Latoszek, Toshihiko Iwahashi, Mio Iwahashi, Shinobu Iwaki, Chieri Kato, Misao Yoshida, Hisanori Sasai, Akira Miyauchi, Naoki Matsushiro, Hidenori Inohara, Makoto Ogawa, Youri Maryn
ObjectivesWe aimed to determine the most appropriate syllable number for analyzing the Acoustic Voice Quality Index for the Japanese-speaking population (AVQIv3-JP) and to validate AVQIv3-JP using the determined syllable number.MethodsFirst, we counted how many syllables should be included in each continuous speech (CS) sample to achieve time-balanced analysis between CS and sustained vowel samples using our previous dataset including 336 CS samples with 58 syllables. From the descriptive statistics of the counted syllable numbers, the most appropriate syllable number was identified. Subsequently, we performed validation procedures of AVQIv3-JP using our latest dataset including 455 recordings.ResultsThirty Japanese syllables were judged to be the most appropriate syllable number. The concurrent validity of the AVQIv3-JP using 30 syllables was confirmed by Spearman's rho of 0.873. Subsequently, the receiver operating characteristic analysis demonstrated the excellent discriminative capability of AVQIv3-JP, showing the area under the curve of 0.915. The AVQIv3's original threshold of 2.43 in the Dutch language corresponded to sensitivity and specificity of 64.6% and 97.3%, respectively. In the present study, a threshold of 1.41 achieved the best accuracy with balanced sensitivity and specificity of 84.4% and 85.6%, respectively. Furthermore, the 95th percentile of the control participants exhibited a threshold of 2.06, showing sensitivity and specificity of 72.1% and 93.8%, respectively, as well as reasonable positive and negative likelihood ratios of 11.7 and 0.298, respectively.ConclusionThe AVQIv3 using 30 Japanese syllables is a reliable measurement tool for estimating the severity of voice quality and detecting abnormal voices.



from #Audiology via xlomafota13 on Inoreader http://ift.tt/2jxIOFC
via IFTTT