嗓音主观听感知评估稳定性的研究

目的 通过对嗓音质量主观听感知评估方法的标准化研究,探讨影响听评委评估结果稳定性的因素.方法 采用符合国际标准的录音设备对300例嗓音障碍患者和100例嗓音正常者进行录音,声音样本为话语声,6位来自不同医院的嗓音医学专家担任听评委,采用GRBAS嗓音评估系统中的总嘶哑度(grade,G)、粗糙声(rough,R)和气息声(breathy,B)作为评估参数.每位评委对3次不同排序嗓音样本的评估采用视觉模拟量表,结果取3次评估的平均值.采用改良分级量表将嗓音质量分为4级水平(0级正常,1级轻度障碍,2级中度障碍,3级重度障碍).采用κ值对评估结果的一致性进行统计学分析,采用回归分析探讨嗓音障碍程度对听评委评估稳定性的影响.结果 不管是评委之间还是评委自身都存在着评估结果的不一致性.评委自身的一致性均达到比较好、好、甚至非常好的水平,其中对参数G评估的一致性最好(κ值0.46～0.85),其次是参数R(κ值0.41～0.84),参数B相对较差(κ值0.41～0.81).评委之间的一致性相对较差,除去1位评委的一致性未达到要求而被剔除外,其他5位评委之间对参数G评估的一致性最好(κ值0.43～0.96),其次是参数R(κ值0.33～0.78),参数B的一致性最差(κ值0.002～0.45).听评委对正常嗓音和重度障碍评估的稳定性高于轻度障碍和中度障碍.结论 影响嗓音主观听感知评估稳定性的最主要因素是听评委之间的不一致性,不同评估参数和嗓音障碍程度影响听评委评估的稳定性.

Abstract：

Objective To explore the factors that influence the stability of evaluation results judged by a jury through a standard research on perceptual evaluation measurements of voice quality. Methods Voice samples from 300 patients with dysphonia and 100 control subjects with normal voice were recorded and assessed by a jury composed of 6 experienced listeners from different hospitals. The voice samples were discourse voices and ordered randomly 3 times, and the mean of 3 evaluations using visual analogue scale were the final results. The jury was instructed to classify voice samples according to the G ( grade) , R (rough) and B(breathy) components of the GRBAS scale on a 4-point scale ranging from 0 for normal to 3 for severe dysphonia. K value was used to analyze the concordance of evaluation results and regression analysis was used to research the effects of the extent of voice disorder to the stability of perceptual evaluation. Results The discordance of evaluation existed both between the jury and in listeners themselves. The concordance of listeners themselves of each evaluation parameter was not bad, good, or even very good, and the concordance of evaluation of G was the best (κ value: 0. 46 - 0. 85) , then R (κ value; 0.41 -0.84) and B(κ value: 0.41 -0.81). The concordance between the jury was worse than that in themselves. And except a listener whose concordance of evaluation was under the requirement, the concordance of evaluation of G was the best ( κ value: 0. 43 - 0. 96) , then R ( κ value: 0. 33 - 0. 78 ) and B ( κ value: 0. 002 - 0. 45 ). The stability of evaluation of normal voice and severe voice disorder was better than mild and moderate voice disorder. Conclusions The discordance between the jury was the main factor that influence the stability of perceptual evaluation. The evaluation parameters and extent of voice disorder will influence the stability of perceptual evaluation of the jury.

Study on the stability of perceptual evaluation of voice quality

Objective To explore the factors that influence the stability of evaluation results judged by a jury through a standard research on perceptual evaluation measurements of voice quality. Methods Voice samples from 300 patients with dysphonia and 100 control subjects with normal voice were recorded and assessed by a jury composed of 6 experienced listeners from different hospitals. The voice samples were discourse voices and ordered randomly 3 times, and the mean of 3 evaluations using visual analogue scale were the final results. The jury was instructed to classify voice samples according to the G ( grade) , R (rough) and B(breathy) components of the GRBAS scale on a 4-point scale ranging from 0 for normal to 3 for severe dysphonia. K value was used to analyze the concordance of evaluation results and regression analysis was used to research the effects of the extent of voice disorder to the stability of perceptual evaluation. Results The discordance of evaluation existed both between the jury and in listeners themselves. The concordance of listeners themselves of each evaluation parameter was not bad, good, or even very good, and the concordance of evaluation of G was the best (κ value: 0. 46 - 0. 85) , then R (κ value; 0.41 -0.84) and B(κ value: 0.41 -0.81). The concordance between the jury was worse than that in themselves. And except a listener whose concordance of evaluation was under the requirement, the concordance of evaluation of G was the best ( κ value: 0. 43 - 0. 96) , then R ( κ value: 0. 33 - 0. 78 ) and B ( κ value: 0. 002 - 0. 45 ). The stability of evaluation of normal voice and severe voice disorder was better than mild and moderate voice disorder. Conclusions The discordance between the jury was the main factor that influence the stability of perceptual evaluation. The evaluation parameters and extent of voice disorder will influence the stability of perceptual evaluation of the jury.