Cepstral Acoustic Analysis of Speech in Children With Prelingual Hearing Impairment Using Hearing Aids and Cochlear
Hamidreza Farmani1, Jaleh Yousefi2, Mahdieh Hasanalifard1
1New Hearing Technologies Research Center, Clinical Sciences Institute, Baqiyatallah University of Medical Sciences, Tehran, Iran.
Background:
Prelingually hearing-impaired children often exhibit atypical voice characteristics due to limited auditory feedback. While time-based acoustic measures (eg, jitter, shimmer) provide useful insights, they have limited reliability and applicability in connected speech. Cepstral analyses, including cepstral peak prominence (CPP) and smoothed CPP (CPPS), offer more consistent and perceptually valid voice assessments. The aim of this study is to compare cepstral acoustic features among children with prelingual hearing impairment using hearing aids (HA) or cochlear implants (CI), and their normal-hearing (NH) peers.
Materials And Methods:
Seventy-two Persian-speaking children (aged 5:6 to 7:11 years old) were divided into three groups (n = 24 each): HA, CI, and NH. Voice samples were collected using the Persian version of the Consensus Auditory-Perceptual Evaluation of Voice and analyzed with PRAAT software to extract fundamental frequency, CPP, and CPPS values from sustained vowels and six sentence types. Data were analyzed using one-way ANOVA and Tukey HSD post hoc tests.
Results:
The HA group showed significantly higher fundamental frequency and lower CPP/CPPS values compared with the CI and NH groups. The CI group demonstrated values more similar to NH peers, particularly in sustained vowel tasks. CPPS measurements were more sensitive in distinguishing group differences than CPP, especially in connected speech. No significant differences were found in voiceless-plosive-weighted tasks.
Conclusion:
Cepstral analyses, particularly CPPS, effectively differentiate voice quality in hearing-impaired children. The findings also suggest that CI users' speech is acoustically closer to that of NH children than to HA users. Moreover, cepstral measures from sentence-based tasks appear more sensitive to speech control difficulties than those from sustained vowel production.
