On soft biometrics

Infosphere, Datafication, and Decision-Making Processes in the AI Era

A recent interpretation of artificial intelligence (AI) (Floridi 2013, 2022) suggests that the implementation of AI demands the investigation of the binding conditions that make it possible to build and integrate artifacts into our lived world. Such artifacts can successfully interact with the world because our environment has been designed to be compatible with intelligent machines (such as robots). As the use of AI becomes ubiquitous in society, possibly leading to the formation of increasingly intelligent bio-technological unions, there will likely be a coexistence of a plethora of micro-environments wrapped and tailored around humans and basic robots. The key element of this pervasive process will be the capacity to integrate biological realms in an infosphere suitable for the implementation of AI technologies. This process will require extensive datafication. This is because data is the basis of the logical-mathematical codes and models that drive and guide AI. This process will have huge consequences on workplaces, on workers, as well as on the decision-making processes required for the functioning of future societies. In this paper we will offer a comprehensive reflection on the moral and social implications of datafication as well as a set of considerations about its desirability, which will be informed by the following insights: (1) full protection of privacy may become structurally impossible, thus leading to undesirable forms of political and social control; (2) worker's freedom may be reduced; (3) human creativity, imagination, and even divergence from AI logic might be channeled and possibly discouraged; (4) there will likely be a push towards efficiency and instrumental reason, which will become preeminent in production lines as well as in society.

Pre-set estimation-based in-silico silhouette-based methodology for improving the robustness to viewing direction difference for assisting forensic gait analysis

Forensic gait analysis is used to visually and quantitatively analyze information regarding the appearance and style of walking that can be presented as evidence in the court. The demand for analyzing CCTV pedestrian footage in video surveillance has been increasing. The dependence of the accuracy of semiautomatic silhouette-based analysis, often used in forensic science, on the differences in the viewing directions is a very challenging issue that is yet to be resolved for real case applications. Currently, the different viewing directions used in comparison footage significantly decrease the accuracy of same person analysis when using the silhouette-based method, often used in the Japanese forensic science domain. A calibration-based method was previously prosed to resolve this problem, but it requires performing an elaborate measurement procedure at the camera installation site for an accurate analysis. In this study, we propose a novel in-silico silhouette-based analysis method that significantly expands the number of viewing direction pre-set settings to 900 from the 24 used in the previous method. Several software tools have been developed to ensure that all the procedures can be executed on a computer. The experimental results confirm that the accuracy of the proposed method is comparable to that of the calibration-based method. Furthermore, the practical comparison results from actual consultation confirmed the effectiveness of the proposed method under existing viewing direction differences. We therefore anticipate that the proposed method will be beneficial for improving the analysis accuracy in real cases and therefore serve as a substitute of the previous method.

PrivacyNet: Semi-Adversarial Networks for Multi-attribute Face Privacy

Recent research has established the possibility of deducing soft-biometric attributes such as age, gender and race from an individual's face image with high accuracy. However, this raises privacy concerns, especially when face images collected for biometric recognition purposes are used for attribute analysis without the person's consent. To address this problem, we develop a technique for imparting soft biometric privacy to face images via an image perturbation methodology. The image perturbation is undertaken using a GAN-based Semi-Adversarial Network (SAN) - referred to as PrivacyNet - that modifies an input face image such that it can be used by a face matcher for matching purposes but cannot be reliably used by an attribute classifier. Further, PrivacyNet allows a person to choose specific attributes that have to be obfuscated in the input face images (e.g., age and race), while allowing for other types of attributes to be extracted (e.g., gender). Extensive experiments using multiple face matchers, multiple age/gender/race classifiers, and multiple face datasets demonstrate the generalizability of the proposed multi-attribute privacy enhancing method across multiple face and attribute classifiers.

Super-Fine Attributes with Crowd Prototyping

Recognising human attributes from surveillance footage is widely studied for attribute-based re-identification. However, most works assume coarse, expertly-defined categories, ineffective in describing challenging images. Such brittle representations are limited in descriminitive power and hamper the efficacy of learnt estimators. We aim to discover more relevant and precise subject descriptions, improving image retrieval and closing the semantic gap. Inspired by fine-grained and relative attributes, we introduce super-fine attributes, which now describe multiple, integral concepts of a single trait as multi-dimensional perceptual coordinates. Crowd prototyping facilitates efficient crowdsourcing of super-fine labels by pre-discovering salient perceptual concepts for prototype matching. We re-annotate gender, age and ethnicity traits from PETA, a highly diverse (19K instances, 8.7K identities) amalgamation of 10 re-id datasets including VIPER, CUHK and TownCentre. Employing joint attribute regression with the ResNet-152 CNN, we demonstrate substantially improved ranked retrieval performance with super-fine attributes in comparison to conventional binary labels, reporting up to a 11.2 and 14.8 percent mAP improvement for gender and age, further surpassed by ethnicity. We also find our 3 super-fine traits to outperform 35 binary attributes by 6.5 percent mAP for subject retrieval in a challenging zero-shot identification scenario.

Demographic Analysis from Biometric Data: Achievements, Challenges, and New Frontiers

Biometrics is the technique of automatically recognizing individuals based on their biological or behavioral characteristics. Various biometric traits have been introduced and widely investigated, including fingerprint, iris, face, voice, palmprint, gait and so forth. Apart from identity, biometric data may convey various other personal information, covering affect, age, gender, race, accent, handedness, height, weight, etc. Among these, analysis of demographics (age, gender, and race) has received tremendous attention owing to its wide real-world applications, with significant efforts devoted and great progress achieved. This survey first presents biometric demographic analysis from the standpoint of human perception, then provides a comprehensive overview of state-of-the-art advances in automated estimation from both academia and industry. Despite these advances, a number of challenging issues continue to inhibit its full potential. We second discuss these open problems, and finally provide an outlook into the future of this very active field of research by sharing some promising opportunities.

Lightweight Biometric Sensing for Walker Classification Using Narrowband RF Links

This article proposes a lightweight biometric sensing system using ubiquitous narrowband radio frequency (RF) links for path-dependent walker classification. The fluctuated received signal strength (RSS) sequence generated by human motion is used for feature representation. To capture the most discriminative characteristics of individuals, a three-layer RF sensing network is organized for building multiple sampling links at the most common heights of upper limbs, thighs, and lower legs. The optimal parameters of sensing configuration, such as the height of link location and number of fused links, are investigated to improve sensory data distinctions among subjects, and the experimental results suggest that the synergistic sensing by using multiple links can contribute a better performance. This is the new consideration of using RF links in building a biometric sensing system. In addition, two types of classification methods involving vector quantization (VQ) and hidden Markov models (HMMs) are developed and compared for closed-set walker recognition and verification. Experimental studies in indoor line-of-sight (LOS) and non-line-of-sight (NLOS) scenarios are conducted to validate the proposed method.