Jh. Walton et Rf. Orlikoff, SPEAKER RACE IDENTIFICATION FROM ACOUSTIC CUES IN THE VOCAL SIGNAL, Journal of speech and hearing research, 37(4), 1994, pp. 738-745
One-second acoustic samples were extracted from the mid-portion of sus
tained /a/ vowels produced by 50 black and 50 white adult males. Each
vowel sample from a black subject was randomly paired with a sample fr
om a white subject. From the tape-recorded samples alone, both expert
and naive listeners could determine the race of the speaker with 60% a
ccuracy. The accuracy of race identification was independent of the li
stener's own race, sex, or listening experience. An acoustic analysis
of the samples revealed that, although within ranges reported by previ
ous studies of normal voices, the black speakers had greater frequency
perturbation, significantly greater amplitude perturbation, and a sig
nificantly lower harmonics-to-noise ratio than did the white speakers.
The listeners were most successful in distinguishing voice pairs when
the differences in vocal perturbation and additive noise were greates
t and were least successful when such differences were minimal or abse
nt. Because there were no significant differences in the mean fundamen
tal frequency or formant structure of the voice samples, it is likely
that the listeners relied on differences in spectral noise to discrimi
nate the black and white speakers.