Facial attractiveness for enhanced face recognition: a novel soft biometric trait

Abstract

This research introduces facial attractiveness as a significant new feature in soft biometrics, aimed at enhancing the accuracy of automated facial recognition systems. Despite its potential to improve recognition systems, facial attractiveness, unlike other standard features such as age, gender, and skin colour, has not received extensive research attention. This study addresses this gap by demonstrating that facial attractiveness can serve as an additional and valuable attribute for identifying individuals.
This research employs a comparative analysis approach to measure facial attractiveness through a structured evaluation method that combines soft biometric data with machine learning techniques. The method involves collecting and ranking attractiveness attributes using the Elo rating system, which are then integrated into recognition models. Experiments indicate that facial attractiveness improves recognition performance, proving its usefulness in automated systems.
Additionally, this research examines the psychological and social aspects of facial attractiveness and considers how they can influence the functionality of automated systems. It discusses the challenges of measuring attractiveness consistently across different datasets, providing a clear overview of the limitations.
Notably, the findings demonstrate that models using soft-biometric attributes alone, including attractiveness, consistently outperformed systems that combined these attributes with Eigenface representations. This outcome underscores the discriminative strength of soft biometrics as standalone inputs rather than supplementary enhancements.
Although this study does not use deep learning techniques, it establishes a foundation for future research by proposing an innovative approach to incorporating facial attractiveness into biometric systems. The author avers that this is the first study to thoroughly explore facial attractiveness as a soft biometric feature.

More information

Identifiers

ORCID for Moneera ALNAMNAKANI: orcid.org/0009-0000-8266-3534

ORCID for Mark Nixon: orcid.org/0000-0002-9174-5934

Catalogue record

Export record