Musician effect in cochlear implant simulated gender categorization

Abstract

Musicians have been shown to better perceive pitch and timbre cues in speech and music, compared to non-musicians. It is unclear whether this "musician advantage" persists under conditions of spectro-temporal degradation, as experienced by cochlear-implant (CI) users. In this study, gender categorization was measured in normal-hearing musicians and non-musicians listening to acoustic CI simulations. Recordings of Dutch words were synthesized to systematically vary fundamental frequency, vocal-tract length, or both to create voices from the female source talker to a synthesized male talker. Results showed an overall musician effect, mainly due to musicians weighting fundamental frequency more than non-musicians in CI simulations.

FIG. 1.

Power spectrum and waveform of the vowel /u/ in “bus”. Each row represents a different voice. The middle row shows the stimulus resynthesized, in STRAIGHT, with the original parameters of the recorded female voice. In the top row, only the F0 was changed, by an octave down. In the bottom row, only the VTL was changed to be made 23% longer, which results in shifting all the formants down by 3.6 st. The left panel shows the spectra over the duration of the vowel, for the vocoded (right column) and non-vocoded (left column, noted “Original”) versions of the stimulus. The black solid line represents the spectrum itself, making the harmonics and/or the sinusoidal carriers (and sidebands) of the vocoder visible. The dashed gray line represents the spectral envelope, as extracted by STRAIGHT on the left, and interpolating between the carriers for the vocoded sounds on the right. The triangles and stems point to the location of the first three formants, as defined by visual inspection of the STRAIGHT envelope, both for the left and right columns. In the right column, the vocoder analysis filter bands are shown with grayed areas. The frequency of the sine-wave carrier is marked with a dotted line.

FIG. 2.

Mean gender categorization (across subjects and test words) shown for non-musicians (left panels) and musicians (right panels) tested with unprocessed (but synthesized) stimuli, as a function of the difference in F0 (top plots) or VTL (bottom plots) relative to the female source talker. Error bars denote one standard error of the mean.

FIG. 3.

Mean gender categorization (across subjects and test words) shown for non-musicians (left panels) and musicians (right panels) tested with the CI simulation, as a function of the difference in F0 (top plots) or VTL (bottom plots) relative to the female source talker. Error bars denote one error of the mean.