Variation in voice quality within speakers
Yoonjeong Lee, Jody E. Kreiman · The Journal of the Acoustical Society of America · 2019
Little is known about the nature or extent of everyday variability in voice quality within a speaker or how this differs across speakers. Using a suite of measures that map between acoustics and perception of voice quality, this study elucidates which acoustic variables within speakers’ individual voice spaces best characterize speakers. Based on studies of faces and cognitive models of speaker recognition, we hypothesized that a few measures would be important across speakers, but that much intra-speaker variability would be idiosyncratic. By using principal component analysis, we tested this hypothesis against a set of multiple sentence productions from 100 native speakers of English (fifty females and fifty males), recorded over three days. Acoustic variables measured every 5 ms on vowels and approximants corresponded to F0, vowel quality, spectral noise, source spectral shape, and variability. Across speakers the balance between higher harmonic amplitudes and inharmonic energy in the voice accounted for the most variance (females = 20%/males = 22%). Vowel quality and its variability accounted for an additional 12%/12% of variance. Remaining variance appeared largely idiosyncratic, suggesting that the speaker-specific voice space is different for different people. Notably, F0 did not emerge from these analyses. Implications for voice recognition are discussed. [Work supported by NIH/NSF.]