Creation, perception, and use of gender expansive synthetic voices

Abstract

Gender expansive (transgender and nonbinary) listeners are typically left out of the study of sociophonetic perception, and even phonetic studies more broadly, although there is an increasing number of studies investigating gender expansive speech production. Various studies have been conducted on the speech of men and women, from which conclusions have been drawn about gender cues in speech. However, these prior investigations are restrictive because they largely consider only cisheterosexual men and women. Research of gender in speech has also impacted the available types of gendered synthetic voices for those relying on them, and little has been documented about the perception of gender in synthetic voices used by Speech Generating Device (SGD) users. This dissertation provides a series of four studies for broadening our conception of gender production and perception in speech and extending this domain beyond the “biological” voice, that is, the attributes of voices that can be traced solely and directly to anatomical or physiological factors. First, the series evaluates the speech of 16 gender expansive participants evaluating f0, formant frequencies, and spectral qualities of [s]. Second, the series evaluates perception of synthetic voices constructed from the same talkers. Finally, the series investigates, in a detailed case study, how a nonbinary SGD user encodes their gender using the aforenoted gender expansive synthetic voices. The results from this dissertation series will enrich the respective fields of sociophonetics, psycholinguistics, and speech technology.