Face Matching Research Articles

Emotion knowledge in neurotypical children and in those with down syndrome

This research aimed to assess two components of emotion knowledge (EK): receptive EK with face emotion identification and matching tasks, and emotion situation knowledge with the emotion attribution task (EAT). Study 1 assessed the development of EK in 265 neurotypical (NT) children (4–11 years), divided into four age groups. Overall, results showed a significant improvement of EK with age in the NT population for the three tasks, especially between the ages of 4/5 and 6/7. Children were less successful at the EAT in comparison to the other two tasks, indicating that receptive EK develops earlier than emotion situation knowledge. The presence of visual context (EAT) does not help to improve our children’s overall facial emotion recognition, especially for anger and sadness, while these emotions are well recognized in isolated facial expressions (emotion identification). Study 2 compared EK between 32 children with Down syndrome (CA: M = 13 years, SD = 2.13) and 32 NT children (CA: M = 5.3 years, SD = 1.36): matched on a vocabulary task. Children with DS had more difficulties in EK than NT children. They had lower performances on the identification and the EAT tasks, while exhibited similar performances to their NT controls on the emotion matching task. Moreover, good abilities to identify emotion expressions seem to be a prerequisite for successful face-context recognition in NT children, but not in children with DS. Difficulties encountered by children with DS could result from executive dysfunction when dealing with complex visual information in addition to emotion processing difficulties.

Patrolling the Borders: Experience with Multiple Out-Group Members is Associated with the Own-Ethnicity Effect

The own-ethnicity effect (OEE) is a recognition deficit for faces of other ethnic groups compared to one’s own group. Thirty-two border patrol (i.e., police) officers at a major international airport expected to have high frequency contact with multiple other ethnic group faces were compared with 32 bank employees and 64 students. German participants viewed out-group faces of African Americans, Mexican Americans/Mexican Nationals, and Turkish faces and German (in-group) faces in a standard recognition and a delayed-matching task. Generally, performance was better as a function of self-reported daily job contact for both tasks. Patrol officers were better with Black faces, while students were better with German faces for the standard recognition task. Although an OEE was present for all groups, it was attenuated for border patrol officers. Job contact with multiple out-groups was associated with face processing. Errors in recognition and matching faces of other ethnic groups pose problems for security in an increasingly global society. Customs agencies and security organizations employing an ethnically diverse workforce along with perceptual training for recognizing and matching out-group members will be better prepared to meet the demands of international security.

Facial plastic surgery and face recognition algorithms: Interaction and challenges. A scoping review and future directions

IntroductionFace recognition (FR) technology can be used in wide range of applications such as identity authentication, access control, and surveillance. Interests and research activities in face recognition have increased significantly over the past twenty years. Plastic surgery procedures can significantly alter facial appearance, thereby posing a serious challenge even to the state-of-the-art face matching algorithms. The purpose of this work was to detail the interaction between facial plastic surgery and facial recognition software and discuss the new challenges of this interaction. Material and methodsThe authors critically reviewed the literature from January 2000 to September 2019, to identify articles reporting interactions between facial plastic surgery and facial recognition algorithms and discuss the new challenges of these interactions. Controlled vocabulary terms and keywords were used in the search strategy and two authors independently analyzed data. Factors included in the analysis were: Author, Journal, Year, Scope, Study design, Plastic surgery, Data (volume, origin and processing), Identification accuracy and Conclusion (interaction/challenge). ResultsForty-three research articles underwent data extraction and 28 articles were included in quantitative synthesis. Of the 28 articles, the most common study designs were experimental evaluation (n=15, 53,5%), Evaluation Study studies (n=7, 25%) and Review studies (n=4, 14,3%). Fifty percent of the articles have been published in the last 4 years (14 articles, 50%). Most of the study scope was informatics (64,3%). Only 10 articles were published in medical journals. Rhytidectomy (face lift) is the most challenging procedure for the FR algorithms. Data volume varied from 4 to 2878 subjects. The proposed algorithms provide at least 15 to 99% better identification performance. Among these, only two papers discuss the new challenges of the interaction between facial plastic surgery and Face Recognition Algorithms. ConclusionIn the context of advances in artificial intelligence, Internet connectivity and data integration, the purpose of this review is, to look forward to analyze the new interactions of facial plastic surgery and facial recognition algorithms, and to suggest avenues for future research and clinical application of this technology. Furthermore, to evaluate if plastic surgeons are prepared to discuss this technology with their patients. Plastic surgeons should be prepared to answer questions from patients about the fundamentals of facial recognition technology, and the potential effects of plastic surgery on facial recognition technology performance. Continued efforts are needed to provide scientifically rigorous data of facial biometric identification after facial plastic surgery and to include these notions in the routine consultation or consent process for patients seeking aesthetic facial surgery.

The McGurk effect in self-recognition of people with schizophrenia

The McGurk effect is a robust illusion phenomenon in the perception of speech; however, there is little research on its demonstration in nonverbal domains. Thus, we tested for the McGurk effect in the context of self-recognition. We presented a group of people with schizophrenia and a control group of people without mental illnesses, with 2 videos accompanied by a soundtrack featuring different identity information. The first video had a matched face and voice; the other featured conflicting face–voice information. The participants judged if the voice in the video was their own or someone else's. The results show there was a robust McGurk effect in self-recognition, which was stronger among participants with schizophrenia because of the influence of self-disorder. Further, people with schizophrenia were less accurate in voice self-recognition when there was conflicting face–voice identity information. Thus, presenting audiovisual-consistent information is conducive to information processing for people with schizophrenia.

Online Deadline-Aware Task Dispatching and Scheduling in Edge Computing

In this article, we study online deadline-aware task dispatching and scheduling in edge computing. We jointly considerthe management of the networking and computing resources to meet the maximum number of deadlines. We propose an online algorithm, named Dedas, which greedily schedules newly arriving tasks and considers whether to replace some existing tasks in order to make the new deadlines satisfied. We derive a non-trivial competitive ratio of Dedas theoretically, and our analysis is asymptotically tight. Besides, we implement a distributed approximation D - Dedas with a better scalability and less than 10 percent performance loss compared with the centralized algorithm Dedas. We then build DeEdge, an edge computing testbed installed with typical latency-sensitive applications such as IoT sensor monitoring and face matching. We adopt a real-world data trace from the Google cluster for large-scale emulations. Extensive testbed experiments and simulations demonstrate that the deadline miss ratio of Dedas is stable for online tasks, which is reduced by up to 60 percent compared with state-of-the-art methods. Moreover, Dedas performs well in minimizing the average task completion time.

3D face reconstruction from mugshots: Application to arbitrary view face recognition

Mugshots while routinely acquired by law enforcement agencies are under utilized by automated face recognition systems. In this paper, we propose a regression based approach to reconstruct textured full 3D face models from multi-view mugshot images. Using landmarks from the input frontal and profile mugshots of a subject, our method reconstructs his⧹her 3D face shape via either linear or nonlinear regressors. The texture of the mugshot images is mapped to the reconstructed 3D face shape via an efficient seamless texture recovery scheme. Compared with existing 3D face reconstruction methods, the proposed method more effectively utilizes the three-view mugshot face images collected during booking. The reconstructed 3D faces are used to generate realistic multi-view face images to enlarge the gallery and facilitate arbitrary view face recognition. Evaluation experiments have been done on BFM and Bosphorus databases in terms of reconstruction accuracy, and on Multi-PIE and Color FERET databases in terms of recognition accuracy. The results show that the proposed method can reduce the 3D face reconstruction error of the best competitive method from 2.31 mm to 1.88 mm, and improve the recognition accuracy of state-of-the-art deep learning based face matchers by as much as ~4% on Multi-PIE and ~2% on Color FERET despite the high baseline set by them.

A Perancangan Sistem Pengenalan Wajah Secara Real-Time pada CCTV dengan Metode Eigenface:

The development of times and curiosity in a condition become a reason for people to continue to develop security systems at home, one of which is by CCTV. Basically, CCTV security systems only function as recording devices on the scene. Therefore, the security level of the CCTV is still low. For that we need a system that can be a security solution. The system can detect objects in the form of faces as image input. To insert image objects into the system, the system requires a camera. The object detected by the camera will do a matching face with the face image contained in the dataset class. The system is the application of Computer Vision in the security system. Brain memory will provide a picture of a face that we have known before. The analogy can be likened to a machine or device that has the same ability as humans to recognize individuals through facial images. Through this research a comparison of facial image recognition with eigenface algorithm using feature extraction, PCA and LDA will be implemented on a real-time computer platform. The library used in Eigenface is OpenCV. The purpose of this study is to find out which method has a high degree of accuracy in performing facial image recognition by comparing between the two methods used. The problem faced by the author when performing accuracy tests is the different light levels between the dataset and the test subject, and changes in attributes such as hair and beard can affect the resulting accuracy. Based on the test results it is known that the accuracy produced by the Eigenface PCA is better than the LDA eigenface. The best accuracy on eigenface was obtained with a PCA combination of 98.06%.

Seleksi Wajah Digital Menggunakan Algoritma Camshift

Real time for digital face database selection using camshift algorithm] Education taken 4-5 years affects physical development. This study uses student digital video data. The recording results are used to identify certain characteristics possessed by a student later stored in the digital file database catalog. The stages of the study consisted of identification, recognition and matching of faces. It starts from converting .mp4 videos to .AVI format. The CAMShift algorithm uses basic HSV colors for tracking face position (tracking) and faces recognition. 1-2 seconds video produces 45-200 frames PNG file. The face matching test results were carried out on several video play, the success of detection: 100% selected, 45%-60%, 80-90%, concluded around 50%-100% successful. Face movements will be caught by the centroid bounding box, if the color of the face is dominant in Hue.

Individual differences in visual acuity and face matching ability

SummaryThe visual acuity of the eyes varies outside the range of normal vision, requiring corrective lenses, but also within the normal range. This study investigated whether both types of variation relate to individual differences in face‐identity matching, considering this applied task requires perception of detail. Across two experiments, face‐matching accuracy correlated with variation in acuity when this fell outside the normal range of vision and was uncorrected with glasses or contact lenses. In contrast, variation in visual acuity within the normal range did not affect face‐matching accuracy, whereas matching accuracy at a given level of acuity could vary substantially. These results indicate that visual acuity is only a problem for occupations performing face‐identity matching when below‐normal acuity is not diagnosed or adequately corrected. In turn, these findings suggest that variation in acuity within the normal range is not a contributing factor to individual differences in face matching accuracy.

Guess who? Facial identity discrimination training improves face memory in typically developing children.

While vast individual differences in face recognition have been observed in adults, very little work has explored when these differences come online during development, their domain specificity, and their consistency across different aspects of face processing. These issues do not only have important theoretical implications for the cognitive and developmental psychological literatures, but may reveal critical windows of neuroplasticity for optimal remediation of face recognition impairments. Here, we describe the first formal remedial face training program that is suitable for children, modifying the popular game Guess Who. Eighty-one typical children Aged 4-11 years were randomly allocated to an experimental or active control training condition. Over 10 training sessions, experimental participants were required to discriminate between faces that differed in feature size or spacing across 10 levels of difficulty, whereas control participants continuously played the standard version of Guess Who within the same timeframe. Improvements in face memory but not face matching were observed in the experimental compared to the control group, but there were no gains on tests of object matching or memory. Face memory gains were maintained in a 1-month follow up, consistent across age, and larger for poorer perceivers. Thus, this study not only presents a promising means of improving face recognition skills in children, but also indicates a consistent period of plasticity that spans early childhood to preadolescence, implying early segregation of face versus object processing. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

A Novel Multisupervised Coupled Metric Learning for Low-Resolution Face Matching

This paper presents a new multisupervised coupled metric learning (MS-CML) method for low-resolution face image matching. While coupled metric learning has achieved good performance in degraded face recognition, most existing coupled metric learning methods only adopt the category label as supervision, which easily leads to changes in the distribution of samples in the coupled space. And the accuracy of degraded image matching is seriously influenced by these changes. To address this problem, we propose an MS-CML method to train the linear and nonlinear metric model, respectively, which can project the different resolution face pairs into the same latent feature space, under which the distance of each positive pair is reduced and that of each negative pair is enlarged. In this work, we defined a novel multisupervised objective function, which consists of a main objective function and an auxiliary objective function. The supervised information of the main objective function is the category label, which plays a major supervisory role. The supervised information of the auxiliary objective function is the distribution relationship of the samples, which plays an auxiliary supervisory role. Under the supervision of category label and distribution information, the learned model can better deal with the intraclass multimodal problem, and the features obtained in the coupled space are more easily matched correctly. Experimental results on three different face datasets validate the efficacy of the proposed method.

Classification of Social Anxiety Disorder With Support Vector Machine Analysis Using Neural Correlates of Social Signals of Threat.

Threatening faces are potent cues in social anxiety disorder (SAD); therefore, neural response to threatening faces, particularly regions in the “fear” circuit such as amygdala, may classify individuals with SAD. Previous studies of indirect/implicit processing of threatening faces have shown that support vector machine (SVM) pattern recognition significantly differentiates individuals with SAD from healthy participants, though evidence for the role of the fear circuit in classification has been inconsistent. We extend this literature by using SVM during direct face processing. Individuals with SAD (n=47) and healthy controls (n=46) completed a validated emotional face matching task during functional MRI, which included a matching shapes control condition. SVM was based on brain response to threat (vs. happy) faces, threat faces (vs. shapes), and threat/happy faces (vs. shapes) in 90 regions encompassing frontal, limbic, parietal, temporal, and occipital systems. Recursive feature elimination (RFE) was used for feature selection and to rank the contribution of regions in predicting SAD diagnosis. SVM results for threat (vs. happy) faces revealed satisfactory accuracy (e.g., area under the curve=0.72); results with shapes as “baseline” yielded less optimal classification. RFE for threat (vs. happy) indicated that all 90 brain regions were necessary for classification. RFE-based ranking suggested diffuse neurofunctional activation to threat (vs. happy) faces in classification. When using an RFE cut-point, regions implicated in sensory and goal-directed processes contributed relatively more in differentiating SAD from controls than other regions. Results suggest that neural activity across large-scale systems, as opposed to fear circuitry alone, may aid in the diagnosis of SAD.

Issues Related to Face Recognition Accuracy Varying Based on Race and Skin Tone

Face recognition technology has recently become controversial over concerns about possible bias due to accuracy varying based on race or skin tone. We explore three important aspects of face recognition technology related to this controversy. Using two different deep convolutional neural network face matchers, we show that for a fixed decision threshold, the African-American image cohort has a higher false match rate (FMR), and the Caucasian cohort has a higher false nonmatch rate. We present an analysis of the impostor distribution designed to test the premise that darker skin tone causes a higher FMR, and find no clear evidence to support this premise. Finally, we explore how using face recognition for one-to-many identification can have a very low false-negative identification rate and still present concerns related to the false-positive identification rate. Both the ArcFace and VGGFace2 matchers and the MORPH dataset used in our experiments are available to the research community so that others should be able to reproduce or reanalyze our results.

The Quantified and Customised Museum

Abstract Mobile and wearable devices provide a range of new tools and approaches to measure the output metrics of the human body, especially in the medical and fitness realms. Cultural institutions similarly are drawing on a range of digital technologies to better understand the neural processes associated with visitor appreciation of artefacts. Harnessing data about how artworks and objects are experienced can be derived from measurable physical observations, such as an individual’s facial geometry, heartbeat, or retina movements. The Peabody Essex Museum in Salam, Massachusetts is used as a case study to explore the onsite quantification of the museum body. This example is placed in conversation with Google’s face match app, which uses computer vision to link user selfies with cultural collections. The discussion focuses on the evolution of a more quantified modality of cultural engagement, which places greater emphasis on the collection of data as an indicator of the quality of a museum visit.

Normative data for two challenging tests of face matching under ecological conditions

BackgroundUnfamiliar face processing is an ability that varies considerably between individuals. Numerous studies have aimed to identify its underlying determinants using controlled experimental procedures. While such tests can isolate variables that influence face processing, they usually involve somewhat unrealistic situations and optimized face images as stimulus material. As a consequence, the extent to which the performance observed under laboratory settings is informative for predicting real-life proficiency remains unclear.ResultsWe present normative data for two ecologically valid but underused tests of face matching: the Yearbook Test (YBT) and the Facial Identity Card Sorting Test (FICST). The YBT (n = 252) measures identity matching across substantial age-related changes in facial appearance, while the FICST (n = 218) assesses the ability to process unfamiliar facial identity despite superficial image variations. To determine the predictive value of both tests, a subsample of our cohort (n = 181) also completed a commonly used test of face recognition and two tests of face perception (the long form of the Cambridge Face Memory Test (CFMT+), the Expertise in Facial Comparison Test (EFCT) and the Person Identification Challenge Test (PICT)).ConclusionsFocusing on the top performers identified independently per test, we made two important observations: 1) YBT and FICST performance can predict CFMT+ scores and vice versa; and 2) EFCT and PICT scores neither reliably predict superior performance in ecologically meaningful and challenging tests of face matching, nor in the most commonly used test of face recognition. These findings emphasize the necessity for using challenging and ecologically relevant, and thus highly sensitive, tasks of unfamiliar face processing to identify high-performing individuals in the normal population.

Hyper-realistic Face Masks in a Live Passport-Checking Task.

Hyper-realistic face masks have been used as disguises in at least one border crossing and in numerous criminal cases. Experimental tests using these masks have shown that viewers accept them as real faces under a range of conditions. Here, we tested mask detection in a live identity verification task. Fifty-four visitors at the London Science Museum viewed a mask wearer at close range (2 m) as part of a mock passport check. They then answered a series of questions designed to assess mask detection, while the masked traveller was still in view. In the identity matching task, 8% of viewers accepted the mask as matching a real photo of someone else, and 82% accepted the match between masked person and masked photo. When asked if there was any reason to detain the traveller, only 13% of viewers mentioned a mask. A further 11% picked disguise from a list of suggested reasons. Even after reading about mask-related fraud, 10% of viewers judged that the traveller was not wearing a mask. Overall, mask detection was poor and was not predicted by unfamiliar face matching performance. We conclude that hyper-realistic face masks could go undetected during live identity checks.

Soft biometrics‐based face image retrieval using improved grey wolf optimisation

Humans often use faces to recognise and identify individuals. Face recognition is one of the important tasks carried out by forensic examiners manually during their investigation, when there is an evidence image/video available from a crime scene. There is a growing demand for face recognition from unconstrained images, which is valuable for criminal investigators in identifying the victims. When an input face image is given to the proposed system, it filters from large scale face dataset to find the top-k similar faces. Deep convolutional neural network approach is employed to extract important features present in the input face image and improved grey wolf optimisation approach is proposed to select optimal features from the extracted features. It is then preceded by a soft biometric-based face matcher that helps in retrieving exact face image from the top-k similar faces matched using approximate nearest neighbour. The performance of the proposed system is evaluated using LFW, CASIA, Multi-pie and Color-Feret datasets. The proposed system addresses the challenge of searching face images from a large collection of unconstrained images by incorporating feature retrieval using DCNN and IGWO with soft biometric face matcher in cascaded framework which improves accuracy and reduces computation and retrieval time.

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.

Amygdala response to emotional faces in adolescents with persistent post-concussion symptoms.

Approximately 30% of adolescents with concussion develop persistent post-concussion symptoms (PPCS) that include emotional symptoms. Elevated amygdalae reactivity to emotional faces has been reported in a variety of psychopathologies characterized by emotional symptoms overlapping with those in PPCS. We tested the hypothesis that amygdalae reactivity to emotional faces in adolescents with PPCS+ is elevated compared to concussed adolescents without PPCS and healthy controls. Concussed adolescents (ages 14-18) with (PPCS+; n=23) and without PPCS (PPCS-; n=13) participated in visits at least 4 weeks post-injury. Adolescents without prior concussion served as controls (HC; n=15). All participants completed a detailed clinical battery and a common emotional face processing task that involved matching of emotional faces or shapes. Compared to HC and PPCS-, adolescents with PPCS+ had elevated depression symptoms, anhedonia, general psychological symptoms, and anxiety symptoms. Contrary to our hypothesis, PPCS+ had lower amygdalae activity to the emotional faces versus shapes condition relative to HC and a trend for lower activity relative to PPCS-. There was a non-significant inverse association between anhedonia amygdalae activity in adolescents with PPCS. Results suggest that adolescents with PPCS have altered amygdalae activity during the processing of emotional face stimuli.

Biometric face classification with the hybridised rough neural network

Biometric face classification is an important indexing scheme to reduce face matching time for large volumes of a database. In this paper, a hybridised approach based on rough set theory (RST) and ...