Divergent HNR and CPP Aging Trajectories in Korean Connected Speech

This study aimed to characterize acoustic aging patterns in Korean connected speech using ecologically valid, tablet-based recordings collected in home environments.

A total of 207 healthy older adults aged 50-81 years (77 male, 130 female; mean age 61.0 years) completed a standardized paragraph reading task, which was recorded using Samsung Galaxy tablets in their home environments.

Multiple linear regression identified distinct age-related trajectories across acoustic measures. Specifically, significant age-related increases were observed in perturbation and harmonic structure, including harmonics-to-noise ratio, jitter RAP, shimmer APQ3, and jitter PPQ5. Conversely, both mean fundamental frequency (F0) and cepstral peak prominence (CPP mean) significantly decreased with age. The age-related F0 decline was driven primarily by female speakers. No significant Age×Gender interactions were detected, although the smaller male subsample limited statistical power to detect such interactions, so this null result does not establish that the trajectories are equivalent. Notably, HNR increased while CPP simultaneously decreased with age, a divergence that contrasts with the sustained vowel literature and that we tentatively interpret as possible compensatory hyperadduction.

Overall, the results show that tablet-based home recordings can yield interpretable age-related acoustic patterns in connected speech and support scalable data collection for acoustic aging research; test-retest reliability and device/environment effects should be quantified in future work.