Perceived Femininity in Singing Voice: Analysis and Prediction

TL;DR

This paper extends the study of perceived gender in voice from speech to singing by introducing Perceived Singing Voice Femininity (PSVF). It builds a stimuli-based survey with 1200 three-second segments across five languages and four singer age groups, collecting 7258 responses from 126 participants and releasing the dataset publicly. The authors define an Average Correspondence (AC) metric to assess how well perceived femininity aligns with singer sex, finding no robust differences across participant demographics but some subgroup trends (e.g., Mandarin tracks and specific age groups). They also adapt the Strada x-vector model to predict a continuous PSVF score, achieving an MAE of 0.10 across five folds, enabling scalable, non-binary analysis of gender biases in music content. Overall, the work provides a data-driven framework to study gender perception in singing and a practical tool for sociological analysis of biases in large music corpora.

Abstract

This paper focuses on the often-overlooked aspect of perceived voice femininity in singing voices. While existing research has examined perceived voice femininity in speech, the same concept has not yet been studied in singing voice. The analysis of gender bias in music content could benefit from such study. To address this gap, we design a stimuli-based survey to measure perceived singing voice femininity (PSVF), and collect responses from 128 participants. Our analysis reveals intriguing insights into how PSVF varies across different demographic groups. Furthermore, we propose an automatic PSVF prediction model by fine-tuning an x-vector model, offering a novel tool for exploring gender stereotypes related to voices in music content analysis beyond binary sex classification. This study contributes to a deeper understanding of the complexities surrounding perceived femininity in singing voices by analyzing survey and proposes an automatic tool for future research.