Acoustic correlates to perceived age in American English.
James D. Harnsberger · The Journal of the Acoustical Society of America · 2009
An acoustic analysis of voice, articulation, and prosodic cues to perceived age was completed for a speech database consisting of 150 speakers of American English. The speakers (75 male, 75 female) were recorded representing three age groups: young (18–30), middle-aged (40–55), and old (62 years and older). The database included several types of read speech: two passages (Rainbow, Grandfather), 16 sentences incorporating a wide phonemic repertoire, three sustained vowels ([a], [i], [u]), and multiple repetitions of nonsense words. Perceptual judgments of age were collected for the sentence materials from 160 listeners in total. These perceived ages were submitted to a multiple linear regression analysis with up to 34 acoustic measures representing: voice quality (e.g., jitter, shimmer, HNR, CPP), articulation (F1+F2 vowel space size, F1 and F2 of point vowels), fundamental frequency (mean, min, max, SD), intensity (SD), speaking rate, and others. An analysis of the sustained vowel data for the females showed that 14 variables were significantly correlated with perceived age (R-squared = 0.78, p < 0.01), while 13 significantly correlated for the male voices (R-squared = 0.57, p < 0.01). Results for the other classes of speech materials will also be reported.