Department of Speech Therapy, School of Rehabilitation, Tehran University of Medical Sciences, Tehran, Iran.
Department of Speech Therapy, School of Rehabilitation, Tehran University of Medical Sciences, Tehran, Iran.
J Voice. 2024 Nov;38(6):1523.e9-1523.e16. doi: 10.1016/j.jvoice.2022.04.008. Epub 2022 May 20.
OBJECTIVES/HYPOTHESIS: The validity of cepstral analysis (Cepstral Peak Prominence [CPP] and Cepstral Peak Prominence-Smoothed [CPPS]) as an indicator of perceptual dysphonia was investigated in the Persian language STUDY DESIGN: Cross-sectional study.
A total of 223 participants (159 with and 64 without dysphonia) uttered vowels /a/ and /i/, six standard sentences, and non-standard connected speech. All vocal samples were perceptually evaluated by three raters on a visual analog scale and put into four groups (normal voice, mild, moderate, and severe perpetual dysphonia). CPP and CPPS of sustained vowel /a/, reading the second standard sentence, and a sentence extracted from non-standard connected speech were established using "Praat" software. Statistical analysis involved a one-way factorial analysis of variance (ANOVA), Kruskal-Wallis H, Kendall's Tau-b correlation, t test, and receiver operating characteristics (ROC) curve.
The results showed that CPP of sustained vowels and reading the standard sentence and CPPS of sustained vowel differed significantly (P < 0.05), except between the normal voice and mild perpetual dysphonia groups (P > 0.05). The CPP of non-standard connected speech, CPPS of reading the standard sentence, and non-standard connected speech differed significantly between all groups (P < 0.05). The mean of cepstral analysis of all tasks, "averaged CPP," and "averaged CPPS" were significantly different between two groups of the normal voice and perceptual dysphonia (P < 0.05). Correlation between the cepstral analysis and the perceptual ratings demonstrated that the correlation coefficients for CPP and CPPS were between 0.4 and 0.6 (P < 0.05). ROC curve analysis revealed that the area under the ROC curve for "averaged CPP" and "averaged CPPS" was greater than 0.8 (P < 0.05). The values of 22.11 and 12.29 were determined as cut-off scores of "averaged CPP" and "averaged CPPS," respectively.
Cepstral analysis was known as useful clinical tool for diagnosis of perpetual dysphonia and determining its severity level in the Persian language.
目的/假设：研究了倒谱分析（Cepstral Peak Prominence [CPP] 和 Cepstral Peak Prominence-Smoothed [CPPS]）作为感知性发声障碍指标的有效性，研究对象为波斯语。
横断面研究。
共有 223 名参与者（159 名有声带障碍，64 名无声带障碍）发出元音/a/和/i/、六个标准句子和非标准连续言语。所有的声音样本都由三名评估者使用视觉模拟量表进行感知评估，并分为四组（正常嗓音、轻度、中度和重度持续性发声障碍）。使用“Praat”软件建立了持续元音/a/、读第二个标准句子和非标准连续言语句子的 CPP 和 CPPS。统计分析包括单因素方差分析（ANOVA）、Kruskal-Wallis H、Kendall's Tau-b 相关、t 检验和受试者工作特征（ROC）曲线。
结果表明，持续元音和读标准句子的 CPP 以及持续元音的 CPPS 有显著差异（P < 0.05），除了正常嗓音组和轻度持续性发声障碍组之间（P > 0.05）。所有组之间非标准连续言语的 CPP、读标准句子的 CPPS 和非标准连续言语的 CPPS 均有显著差异（P < 0.05）。所有任务的平均倒谱分析、“平均 CPP”和“平均 CPPS”在正常嗓音组和感知性发声障碍组之间有显著差异（P < 0.05）。倒谱分析与感知评分的相关性表明，CPP 和 CPPS 的相关系数在 0.4 到 0.6 之间（P < 0.05）。ROC 曲线分析表明，“平均 CPP”和“平均 CPPS”的 ROC 曲线下面积大于 0.8（P < 0.05）。分别确定 22.11 和 12.29 为“平均 CPP”和“平均 CPPS”的截断分数。
在波斯语中，倒谱分析是诊断持续性发声障碍和确定其严重程度的有用临床工具。
