Augmenting Gabor-based Face Recognition with Global Soft Biometrics

Abstract

Though face recognition using traditional (hard) biometrics has attracted massive research interest and received extensive studies, it still confronts degrading variability challenges and notably achieves lower performance compared with other biometric recognition forms like fingerprint and iris. Recently, a number of research studies have been interested in enhancing face recognition performance by all means of supplementary facial biometric traits or other biometric modalities. Thus, soft biometrics have been emerged as a new promising modality of biometrics and highlighted as likely viable and fusible traits for augmenting traditional/hard biometrics. This is due to the expected advantages of soft biometrics over the traditional biometric traits, such as the high collectability and invariance properties. Other than fusing different kinds of traditional traits to augment face recognition, adding soft biometrics to augment various traditional facial traits has yet gained little research attention. Hence, in this research, unlike the majority of existing work, we investigate the viability of global soft face biometrics in supplementing traditional (hard) biometrics and the efficacy of concurrently using absolute and relative descriptions as soft biometrics. We conduct a new soft biometric-based fusion scheme in feature-level for augmenting a traditional Gabor-based face identification/verification in different potential forensic scenarios, considering performance variability evaluation and comparison with the baseline performance of Gabor features in isolation.