Cepstral Peak Prominence Standard Deviation across the Lifespan


Average cepstral peak prominence (CPP) is recommended for acoustic voice analysis. CPP is a cepstral-based measure that evaluates the periodicity of the vocal signal. While average CPP is often the focus, measures like CPP variability (e.g., standard deviation, SD) may provide additional insights. This is particularly useful when CPP falls within the normative range but significant dysphonia is perceived. This study examined CPP SD across multiple child and adult vocal databases, including individuals with voice disorders (Vgroup) and without voice disorders (Cgroup), using both sustained /a/ and the all-voiced sentence “we were away a year ago.” Preliminary analyses in Praat are discussed below.

Pediatric: CPP SD was evaluated for 313 children (Vgroup: average (M) = 7.2 years, 159 male, 80 female; Cgroup: M = 6.6 years, 41 male, 33 female). Average CPP SD values were significantly lower in the Vgroup for the sentence stimuli (t = -4.82, p < 0.001), with this measure having a moderate predictive accuracy (AUC = 0.67). There were no group differences for the sustained vowel stimuli (p =0.60), and the AUC was no greater than chance (0.50).

Adult: CPP SD was evaluated for 256 adults (Vgroup: M = 53.2 years 70 males, 101 females; Cgroup: M = 30.1 years, 19 male, 69 female). Unlike the pediatric sample, average CPP SD values were significantly higher in the Vgroup for the sustained vowel stimuli (t = 4.59, p < 0.001), with this measure having a moderate predictive accuracy (AUC = 0.67). There were no group differences for the sentence stimuli, with the AUC no greater than chance (0.53).

This work is currently ongoing, and the final dataset will include both values from Praat and ADSV (N.B. preliminary analysis show a high correlation between CPP SD values). The presentation will explore potential reasons for increased CPP SD in the adult Vgroup, yet decreased CPP SD in the pediatric Vgroup. Moreover, all results will be analyzed for subject factors that may influence outcomes (e.g., age, recording location). The discussion will provide a comprehensive evaluation of CPP SD across the lifespan, offering insights valuable to clinicians and voice researchers.

Elizabeth
Shaheeh
Anne
Roger
Roseanne
Karen
Linda
Terri
Heller Murray
Awan
Hseu
Nuss
Clark
Zur
Carroll
Giordano