The Relationship Between Cepstral Peak Prominence and Selected Parameters of Dysphonia

Cepstral peak prominence: A comprehensive analysis

“…The figure shows that the dependence between A\ and jitter is much more significant than the dependence between A\ and shimmer or noise. This is consistent with the findings of Murphy [44] and Heman-Ackah et al [4]. Yet, the relation between A\ and shimmer or noise exists, although it is weaker.…”

“…Such a definite statement made by Maryn et al is based on a meta-analysis that considered previous results by Wolfe and Martin [2], Wolfe et al [3], Heman-Ackah et al [4], Halberstam [5], and Eadie and Baylor [6]; yet, later results are also consistent with that assertion, as those published by Awan et al [7], Awan and Roy [8], Maryn et al [9,10], Shueetal. [11], Alpan et al [12], and Petterson et al [13].…”

“…Thus, setting a fixed threshold ((Jthres = 1.5 ms in this case) implies not being able to model the descent of the cepstral envelope equally well for all vocal tracts. It is then coherent that the averaging of CPP values for different articulatory configurations, that is, the computation of CPP from running speech, provides higher correlations with perceived dysphonia [55,5,4] and also measurements that are more robust against changes in utterances, both in length and phonemic content [56]. Yet, one should bear in mind that the calculation of CPP from sustained vowels and from running speech leads to different statistical distributions ofthe obtained measurements [34].…”

“…Wolf and Martin [2] reported that CPP is a discriminant measure for distinguishing strain from other dysphonic voice types (hoarseness and breathiness) but that does not help in further distinctions among voice qualities; yet, the same authors later found that CPP calculated for band-pass filtered voices might be useful for the identification of voice qualities [3]. Similarly, Heman-Ackah et al concluded on the one hand that CPPs correlates with dysphonia severity more than NHR (noise-to-harmonics ratio), APQ (amplitude perturbation quotient), RAP (relative average perturbation) or smoothed PPQ (pitch perturbation quotient) [4,34] and also with breathiness and roughness but, on the other hand, that correlation with roughness is not relevant [4]. A similar conclusion was reached by Awan and Roy [35].…”

“…In order to have an additional insight into the meaning of CPP, some authors have sought for correlations between CPP and other acoustic parameters. Heman-Ackah et al reported greater correlations between CPP and pitch perturbation measures (RAP and sPPQ) than between CPP and measures of noise (NHR) and amplitude perturbation (APQ) [4]. On the contrary, Samlan et al concluded that HNR is correlated to CPP [29].…”