Cues for Perception of Gender in Synthetic Voices and the Role of Identity

Differences in Sibilant Perception between Gender Expansive and Cisgender Individuals

“…A possible explanation for the group difference is that because pitch was held in the neutral range, the cisgender listeners no longer had that cue to rely on for categorization. In the study by Hope and Lilley, 17 it was found that cisgender listeners perceived a female vocal tract voice that had a higher pitch as significantly more feminine than a female vocal tract voice with a neutral pitch, whereas this difference was not significant for the GE listeners. Thus, with pitch neutral, the cisgender listeners perceived female vocal tract voice in ways that were skewed toward a masculine perception of the sibilant (more [ R ]), while the GE listeners already perceived the female vocal tract as feminine even with neutral pitch, and hence more likely to perceive [s] in the acoustic signal.…”

“…has not yet considered synthetic voice a part of the discussion. Hope and Lilley 17,18 were the first to explore a multidimensional, nonbinary voice gender framework of perception which utilized synthetic voice as stimuli. However, that study was limited to subjective voice gender and did not examine how processing of phonemes themselves may be impacted by gender group and community membership.…”

“…In addition, some f 0 -contour models were trained on modified f 0 data such that the speaker mean f 0 would match the global ("neutral") mean f 0 , while preserving the relative f 0 -contours. See Hope and Lilley 17 for more detailed explanation. Three of the resulting voices were used for this experiment:…”

“…This study seeks to examine how GE people categorize sibilants compared to cisgender people when the voices are kept in a "neutral" pitch range using synthetically made voices with varying vocal tract parameters (i.e., formant frequencies and spectral information). We use three different sets of vocal tract parameters based on the results from Hope and Lilley 17,18 which showed that GE people have a distinct "other" gender perception which is correlated with a nonbinary vocal tract created from averaging across male and female vocal tracts. A neutral pitch range was used for two reasons: (1) the results of Hope and Lilley 17,18 were based on stimuli with pitch held in the neutral range and (2) this ensured that pitch would not sway results and obscure the detection of differences in categorization due to vocal tract.…”

“…We use three different sets of vocal tract parameters based on the results from Hope and Lilley 17,18 which showed that GE people have a distinct "other" gender perception which is correlated with a nonbinary vocal tract created from averaging across male and female vocal tracts. A neutral pitch range was used for two reasons: (1) the results of Hope and Lilley 17,18 were based on stimuli with pitch held in the neutral range and (2) this ensured that pitch would not sway results and obscure the detection of differences in categorization due to vocal tract. The f 0 of the stimuli was also within the "neutral range" (145-175 Hz) according to Davies and colleagues.…”