Modeling the relationship between 3D facial features and human binary sex categorization in young Chinese adults

New technologies, such as VR and AR, allow individuals to interact with digital avatars in increasingly realistic ways. As these applications become more widespread, it is important to understand how humans perceive 3D avatars and their facial features. People usually undergo a rapid unconscious cognitive sex categorization process when they begin to interact with new 3D characters. In this study, we used a sex categorization task to investigate how the human brain perceives 3D facial features under different conditions and to uncover how it carries out human binary sex categorization. With 20 young Chinese adult participants in our study, we aimed to reveal how key factors (3D facial shape, texture, and display method) affect human perceptions of binary sex and related facial feature differences. Using regression and PCA reconstruction, we developed a perceived sexual dimorphism model based on the participants' responses. The cosine similarity and difference modeling revealed a clear deviation from biological dimorphism, indicating that human perceptions simplify facial features, primarily focusing on the cheek region. Additionally, a repeated measures ANOVA and post hoc tests showed that dynamic videos with realistic textures yielded the most accurate sex categorization based on facial shape. In contrast, face stimuli lacking texture led to a noticeable male bias, indicating the importance of balanced texture on characters' faces. This work bridges a critical gap regarding human perceptions of 3D characters’ faces, offering valuable insights into creating 3D characters that are perceptually accurate and socially effective.