Face Profile Biometric Systems: An Overview

The complexity of closed-circuit television (CCTV) systems has increased to provide data recordings from various views including the side view. Accordingly, the growing difficulty in identifying individuals under a variety of surveillance conditions has significantly contributed to higher interests in face profile biometric systems. This section of the book aims to look at biometric profiling systems, not just for their potential benefits but also for the errors and concerns they may raise. In this book chapter, we describe the task of face profile biometrics and its importance. All available face profile datasets and their characteristics and statistics are then described. Next the main contributions for the topic of face profile biometrics in the literature are presented. One of the important works is the one presented by Park and Jain with more than 92% recognition rate for the FERET database with 994 subjects. We also present the notion of soft biometrics to discuss why and where we need to deploy soft biometric techniques, and we also explain how soft biometrics features are integrated with traditional biometrics. In a soft biometric framework, data labelling based on categorical and comparative settings is discussed and is then explained why comparative settings are associated with less human errors in the labelling process. We then describe the importance of soft biometrics and its impact on other biometric modalities such as gait and frontal face. Soft biometric in face profile biometrics is then introduced. The results on face profile soft biometric indicate a significant improvement in recognition rate when fused with traditional biometric to present a total 98% recognition rate on XM2VTSDB dataset with 230 subjects.