The Use of Semantic Human Description as a Soft Biometric

Abstract

Abstract-Gait as a biometric has a unique advantage that it can be used when images are acquired at a distance and other biometrics are at too low a resolution to be perceived. In such a situation, there is still information which can be readily perceived by human vision, yet is difficult to extract automatically. We examine how this information can be used to enrich the recognition process. We call these descriptions semantic annotations and investigate their use in biometric scenarios. We outline a group of visually assessable physical traits formulated as a mutually exclusive set of semantic terms; we contend that these traits are usable in soft biometric fusion. An experiment to gather semantic annotations was performed and the most reliable traits are identified using ANOVA. We rate the ability to correctly identify subjects using these semantically prescribed traits, both in isolation as well as in fusion with an automatically derived gait signature.

Analysing comparative soft biometrics from crowdsourced annotations

“…A large number of previous studies focus on biometric fusion techniques to perform identification, by combining ancillary soft biometric information with traditional hard biometrics like gait [1,37,36], face [39,32] or fingerprints [15,16]. However, subjects must first be enrolled into the system in order to match a known hard biometric, which is often unobservable or occluded in real-world CCTV images, presenting a serious limitation for practical application.…”

“…Soft biometrics aims to enable subject identification from physiological semantic attributes and spans a number of disciplines, including; traditional biometric identification [37,36], forensic investigation [24,40], human psychology [19,22] and pedestrian re-identification [41,2,46]. Two recent surveys explore the field in detail, reviewing techniques for demographic estimation from face, body, clothing and other modalities such as gait, fingerprint and iris [30,5].…”

“…We describe 100 subject images captured in a synthetic environment [38], similar to [37,36,39,6]. This enables a coherent appraisal of the value in crowdsourcing such novel annotations and a comparison of recognition performance to [37,36] who use similar image sets.…”

“…From this we discern the degree of subjectivity, confusion and discrepancy of each soft biometric trait when viewed in an exemplary scenario. Furthermore, to ascertain their correspondence to traditional methods, comparative annotation characteristics are differentiated against categorical annotations collected by Samangooei et al [37]. Lastly, a semantic recognition experiment evaluates the discriminative properties of the inferred relative soft biometrics labels and is contrasted to two previous studies in categorical [37] and comparative [36] soft biometric annotation.…”

“…This enables a coherent appraisal of the value in crowdsourcing such novel annotations and a comparison of recognition performance to [37,36] who use similar image sets. We show the effectiveness of the crowdsourcing task and suitability of the soft biometric traits by analysing the distributions of collected annotations and stability of inferred relative measurements.…”