Cepstral Peak Prominence (CPP) Values in Dysphonic and Nondysphonic Hebrew Speakers

Cepstral Peak Prominence (CPP) is considered a valuable acoustic measure in voice assessment. Many factors, such as spoken language, speech tasks, and analysis protocols, may influence CPP values. This study examined CPP values among Hebrew speakers and their association with the five-item GRBAS perceptual scale.

Participants included 630 adult Hebrew speakers, of which 311 were dysphonic (128 men and 183 women, age range 18-88) and 319 were nondysphonic (138 men and 181 women, age range 18-87). CPP values were extracted from recordings of the sustained vowels /a/ and a reading task, using the Praat software, with and without voice activity detection (VAD). Each recording was also perceptually evaluated by an experienced SLP using the GRBAS scale. Group differences were examined, as were correlations between the two CPP measures (with and without VAD) and among the CPP measures and the five GRBAS scales. Sex differences and associations with age were also examined.

Strong correlations were observed between CPP values obtained from sustained vowels and the reading task (0.76 ≤ r ≤ 0.84). Strong correlations were also observed between CPP values that were extracted with and without VAD (0.97 ≤ r ≤ 0.99). Significant group differences in CPP values were found between dysphonic and nondysphonic speakers (P < 0.001) in both sustained vowel and reading tasks, with greater group discrimination obtained using the sustained vowel task (AUC = 0.77). Finally, CPP values significantly negatively correlated with all five GRBAS scales (-0.23 ≤ r ≤ -0.49), with the strongest correlations obtained for the Grade, Breathiness, and Asthenia scales, and the weakest for Strain and Roughness.

This study provides CPP values, with and without VAD, for adult Hebrew speakers and correlates them with the five GRBAS scales. Both CPP measures separate dysphonic from nondysphonic speakers, with stronger discrimination for sustained vowels than for reading. The findings highlight the importance of language-specific data and the use of reproducible protocols for acquiring and clinically implementing CPP measures.