Using neural networks to understand static and dynamic cues in facial expression recognition.

Face recognition models are vulnerable to spoofing of adversarial patches in the physical world. Attackers can enable face recognition models to make false identity judgments by simply pasting a sticker with a special pattern on the face. However, existing attacks lack the ability to transfer to black-box models, and the improvement of transferability is mainly focused on adversarial perturbations based on the p-norm. To further improve the attack performance and transferability, a highly transferable face recognition adversarial patches generation method named as AdvDiffusion is proposed. It first determines the region for adversarial patches generation based on facial gradient maps, and then an image is reconstructed to generate an adversarial patch by adding noise and denoising it with a pre-trained diffusion model. In the denoising, an adversarial loss is used to fine-tune the model and control the image to generate an adversarial patch with spoofing capability. Experiments and analysis show that the adversarial patches generated by the proposed mehtod have good adversarial attack capability on black-box face recognition models in both digital and physical domains, and also have better robustness under the changes of a complex physical environment compared with some state-of-the-art methods. It has great potential application for black-box attacks in the physical domain.

Although prior research has identified population-level trends in facial emotion recognition (FER) in middle childhood, it is unclear whether all children follow a similar developmental trajectory. To address this gap, we used a person-oriented approach to identify qualitatively distinct FER profiles based on accuracy and bias. The sample (N = 3,717, 51% girls, baseline Mage = 8.20, SD = 0.86) came from a Finnish social-emotional learning intervention study, with data collected across three waves (2013-2015). We applied latent profile analysis and random-intercept latent transition analysis to examine FER profiles and their stability across early school years. Moreover, we assessed whether parenting and children's social-emotional adjustment were associated with FER profile membership and transitions. We found five FER profiles: balanced accuracy (14%-26%), sadness bias (34%-38%), positively biased (34%-38%), anger bias (3%-4%), and low accuracy (2%-7%). While the three largest profiles showed moderate stability, children in the anger bias and low accuracy profiles were more likely to transition to other profiles. Compared with the positively biased profile, children in the anger bias profile experienced lower parental warmth and fewer peer problems and showed less prosocial behavior. These findings suggest substantial heterogeneity in children's FER development and link emotion-specific patterns to children's social functioning. Understanding that children may follow distinct developmental trajectories could help design targeted interventions to support children's social skills. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

The role of prior sleep, particularly nighttime sleep, in memory encoding is poorly understood in infants. The present study, therefore, focused on possible associations between nighttime sleep and memory encoding in the first year of life. First, the sleeping behavior of 6-month-old (n=32) and 12-month-old (n=35) infants was assessed for one night using actigraphy and sleep diaries. The following day, infants participated in a visual recognition memory task in the laboratory, which included photographs of female adults showing happy, angry, and neutral facial expressions. Six-month-old infants did not recognize any of the faces, whereas 12-month-old infants recognized the emotional, but not neutral faces, as indicated by novelty preference scores. Correlation and hierarchical regression analyses revealed that sleep efficiency and light sleep predicted novelty preference scores for angry and neutral faces in 6-month-olds, whereas light sleep and co-sleeping predicted novelty preference scores for angry faces in 12-month-olds. These results suggest that specific sleep parameters have distinct and age-specific significance for subsequent memory encoding.

• DemoFace provides a demographically balanced pixelated real face image dataset to mitigate biases in face biometric systems. • The dataset’s structured image–text embedding multimodality supports downstream tasks and facilitates the analysis of model biases in CVFMs. • Through our novel Responsible AI methodology, we developed the dataset and established a set of baselines using SOTA CVFMs to assess the performance and fairness of CVFMs in face biometric tasks. • Through our evaluation using both new and adapted metrics, we revealed inherent bias patterns in several SOTA CVFMs, providing valuable insights to guide future research. Bias and fairness are critical challenges in data-driven computer vision (CV), where limited demographic diversity in training data worsens these challenges. Face biometric (face recognition) systems are core tasks of CV that are highly impacted by these challenges, as existing real-face datasets lack comprehensive demographic representation, whereas current synthetic datasets promote stereotypes. CV Foundation Models (CVFMs) are currently at the forefront of CV applications, including face biometrics, which use global features in multimodal data. However, the scarcity of large-scale, demographic multimodal datasets, such as image-text embeddings for model fine-tuning (or training), limits the fairness in state-of-the-art (SOTA) CVFMs for downstream face biometric tasks. To address these issues, we introduce DemoFace, a balanced demographic face dataset comprising 30,240 pixelated real face images of 672 representative individuals evenly distributed across 48 demographic groups, categorized by ethnicity/race, gender, and age. We gathered images using an API set up from multiple copyright-free public forums. The collected images were then manually filtered, anonymized, and annotated by two independent research groups, and then lightly pixelated for privacy preservation. DemoFace’s image-text embedding multimodality enables fine-tuning (or training) of CVFMs for fairness-focused face biometrics tasks and bias pattern evaluation. Through two empirical studies: face authentication as classification and textual description as token generation, we established baseline scores across ethnicity/race, gender, and age groups. Our baselines identified inherent bias patterns through both new and tailored metrics derived from existing ones, emphasizing the need for more equitable AI models. Here is the Repository: Link

Getting to know you: Gaze behaviours and biographical information in face-name associations.

Network-level disruption effects familiar face processing via cortical nodes connected to occipito-limbic and tempora-frontal networks.

Similarity-aware contrastive learning for face anti-spoofing via frequency enhancement and reconstruction.

An angle-guided bidirectional feature transformation network for multi-frame tilt-angle face recognition

Anatomical features that define facial identity also influence how a face moves during expression. This overlap suggests the visual system might form expectations about facial motion based on shape. However, recent theories propose that facial shape and motion are processed in independent brain pathways. We used reverse correlation to test whether expectations about natural facial movement are shaped by facial structure, and vice versa. Participants judged synthetic dynamic faces with randomly varying shape or motion. We estimated expected speed of facial expression components and compared these across different identities. We also identified shape features used for face recognition and assessed whether these changed with familiar versus unfamiliar motion. Our results show that expectations of natural motion during expression are influenced by face shape, but the shape features used for identification remain stable across motion contexts. We interpret these findings within the independent pathways framework, suggesting a form of motion-from-structure processing.

Mutual sample-center interaction with hard queue mining for face recognition

Alcohol use disorder and emotional processing patterns: Insights from a systematic review.

Sleep plays a crucial role in memory consolidation, particularly for complex, hippocampal-dependent tasks, such as associating faces with names. While numerous studies have examined the impact of sleep on memory, the specific effects on episodic face recognition, perceptual face processing, and the underlying neural mechanisms remain unclear. Thus study was designed to systematically evaluate the role of sleep in face recognition memory, learning, and consolidation in humans. A systematic search of PubMed, Google Scholar, Embase, Scopus, and Web of Science identified English-language, peer-reviewed studies published up to February 5, 2026. The authors independently screened articles, extracted data, and cross-verified results. The review adhered to PRISMA guidelines to maintain methodological rigor and transparency. Across 19 included studies, overnight sleep, post-learning sleep, and targeted naps showed the strongest benefits for face-related learning and memory, particularly in associative and hippocampal-dependent tasks, such as face-name and face-face memory. Slow-wave sleep/N3 and targeted memory reactivation were associated with improved cued face-name recall, while REM sleep and sleep spindles appeared to contribute to implicit face priming, emotional face memory, and adaptive face-learning processes. Sleep deprivation and sleep restriction generally impaired face-memory performance, particularly when tasks required episodic retrieval, associative binding, emotional discrimination, or sustained cognitive control. In contrast, simpler face-recognition or familiarity-based tasks showed weaker and less consistent sleep-related effects, often suggesting protection against forgetting or interference rather than robust memory enhancement. Sleep mainly benefits face learning when tasks require binding, episodic retrieval, emotional processing, or consolidation, while simple familiarity recognition shows weaker effects.

Video-based emotion recognition is an important topic in affective computing, with applications in human-computer interaction, mental health, and multimedia systems. In this work, we propose a two-stage deep learning approach that extracts spatial features from individual frames and leverages temporal consistency across video sequences for facial expression recognition using the RAVDESS dataset. First, videos are split into frames, and a fine-tuned VGG16 CNN extracts discriminative spatial features from each frame. Second, these features are aggregated into sequences and the frame-level predictions are aggregated at the video level using a majority voting strategy to ensure temporal consistency. Our experiments show that the proposed method achieves 93.6% accuracy at the frame level and 98.1% at the video level, outperforming baseline models and remaining competitive with recent state-of-the-art approaches. Temporal aggregation helps reduce misclassifications of subtle emotions, while fine-tuning improves feature extraction. The approach is computationally efficient and provides a solid foundation for future research in multimodal emotion recognition and advanced video-level aggregation.

In the era of digital transformation characterized by the deep integration of artificial intelligence and the Internet of Things, human-machine interaction systems have become ubiquitous in smart architecture and urban security. As the primary security interface, intelligent access control systems face unprecedented challenges. Traditional biometric technologies, such as facial recognition and fingerprint scanning, not only rely heavily on external power sources but also encounter critical limitations regarding privacy risks and environmental sensitivity. To address these issues, this study develops a self-powered bimodal sensor based on a single-electrode triboelectric nanogenerator, providing a low-power, high-security, and multidimensional sensing solution. The core of the sensor lies in its sophisticated functional structural design, featuring a polydimethylsiloxane triboelectric layer patterned with a micro-pyramid array, integrated with a copper foil electrode and a polyethylene terephthalate substrate. The device leverages the coupled effects of contact electrification and electrostatic induction to convert mechanical stimuli into characteristic electrical signals rich in material electronegativity and human kinetic information. To extract latent features from these complex waveforms, a deep learning framework based on a convolutional neural network is implemented to analyze and decouple the signals. Experimental results demonstrate that the system exhibits superior recognition performance in complex environments: in single-dimensional tasks, the accuracies for material identification and user authentication reach 99.83% and 98.88%, respectively. Even under a challenging scenario, the system maintains a high recognition accuracy of 96.15%. This work provides a robust technological foundation for future smart security, flexible electronic skins, and personalized healthcare monitoring.

A novel face scale and National Early Warning Score 2 prediction model to assess deteriorating patients in hospital wards: A quick visual early warning score study.

ABSTRACTObjectiveEffects of obesity on brain health have been revealed in adults, including mental health effects and cognitive impairment. Regarding cognition, obesity‐related memory impairment has been specifically described. While this effect could have a major impact on learning abilities during adolescence, few studies have considered this critical period.MethodsIn this present study, a new fMRI memory task based on paired encoding of faces and backgrounds and subsequent face recognition was presented to male adolescents living with obesity (N = 11) and their lean counterparts (N = 15).ResultsOur study shows that adolescents living with obesity exhibited significantly lower face recognition memory performances (F(1,72) = 9.84, p = 0.002) than their lean counterparts coupled with altered functional cerebral activation patterns during the encoding and retrieval phases of the task. More specifically, during the encoding of the task, a hypoactivation of the right hippocampus and the parahippocampal gyrus was identified in adolescents living with obesity and during the retrieval a hyperactivation of the precuneus (Z > 2.3, cluster‐corrected p = 0.05).ConclusionsThese results suggest that obesity during adolescence is associated with neurocognitive impairment. Future studies should consider adolescence more carefully since this memory impairment could contribute to academic learning difficulties faced by adolescents living with obesity.Trial RegistrationCHUBX 2017/19; CPP number: 2017‐3A02533‐50

Emotion recognition (ER) refers to the perceptual and cognitive processes involved in identifying emotional expressions in others, whereas alexithymia reflects difficulties in identifying and describing one's own emotions. Mood and anxiety symptoms, as well as exposure to stress, have been associated with alterations in emotional processing. This study examined the associations between stress levels, self-emotion processing (alexithymia), and recognition of others' facial expressions in a non-clinical student sample. One hundred twenty-four college students completed questionnaires assessing stress, anxiety (BAI), depression (BDI), and alexithymia (TAS-20). Stress typology measurements included Childhood Trauma Questionnaire (CTQ), Perceived Stress Scale (PSS) and Chronic Stress Scale (CSS). The Emotion Recognition (ER-40) task of PennCNB assessed the ability to recognize angry, fearful, sad, happy, and neutral facial expressions. ER-40 performance was not significantly correlated with stress measures or psychopathology scores, except for the negative correlation of recognition of fear expressions with TAS-20 scores. The relatively high accuracy levels and limited variability observed in task performance may have reduced sensitivity to detect subtle associations, suggesting the possibility of ceiling effects. In contrast, all stress and psychopathology measures were positively correlated with alexithymia scores (BAI: p < 0.001, BDI: p < 0.001, CTQ: p = 0.008, PSS: p < 0.001, CSS: p = 0.001). In linear regression analysis, alexithymia, particularly "difficulty in identifying feelings" subscale (TAS_DIF), was found to be associated with scores in CSS and PSS, when corrected for age, gender and CTQ. Mediation analysis indicated that the association between stress measures and TAS_DIF was statistically associated through anxiety symptoms (p < 0.001), while associations through depressive symptoms were not significant. Higher levels of perceived and chronic stress were associated with greater difficulties in identifying one's own emotions, and these associations were statistically mediated in part by anxiety symptoms. No significant associations were observed between stress and facial emotion recognition performance in this sample, nevertheless, the null findings related to the ER-40 should be interpreted cautiously, particularly in light of potential ceiling effects and the relatively high-functioning nature of the sample. Interventions targeting anxiety and stress management may be beneficial and should be studied for students presenting with elevated stress and alexithymic traits, potentially supporting emotional clarity and adaptive coping in preventive mental health settings.

In item-method directed forgetting, directed forgetting occurs when remember-cued items are better retained than forget-cued items, a result often attributable to the rehearsal of remember-cued items and cessation of rehearsal of forget-cued items. Such effects were found in Experiment 1 for happy White and Asian faces, but not when neutral and sad White or Asian faces were presented at study. These results were consistent with predictions from the affect-as-cognitive-feedback hypothesis, according to which positive stimuli promote task-relevant processing whereas negative stimuli stop or reverse those processes. Results of Experiment 2 suggest that Experiment 1 results are not easily attributed to dimensions associated with the photographs (e.g., attractiveness) used in Experiment 1. In Experiment 3, recognition was examined for photographs of both Whites and Asians seen at study. Consistent with research on face recognition biases and the 'happy-face effect', more White than Asian faces and more happy than sad photographs were recognized at test. Directed forgetting was found when attention was directed to in-group or out-group faces but with increased task complexity or demands, stimulus attributes come to guide information processing. Suggestions for future studies are presented.

The rapid digital transformation of educational institutions necessitates intelligent systems that enhance visitor experience and streamline administrative processes. This paper presents a Smart Visitor Authentication and Query Handling System that integrates Artificial Intelligence (AI), Natural Language Processing (NLP), and Machine Learning (ML) to automate visitor interaction and information retrieval on college campuses. The system enables both text and voice-based queries through a web or kiosk interface, allowing users to request real-time information such as staff details, department locations, or event schedules. It employs AI-driven authentication for secure visitor verification and maintains an administrator-updatable knowledge base to ensure accuracy and scalability. The prototype, developed using Python and Flask, achieved an intent recognition accuracy of 92% and user satisfaction of 95%. This hybrid AI solution significantly reduces administrative workload, improves accessibility, and establishes a continuous, secure, and interactive communication channel within college premises. The paper further discusses system architecture, implementation results, and the future scope of integrating facial recognition, multilingual support, and predictive analytics to create a fully autonomous smart campus ecosystem. Keywords—visitor management system; QR code; campus security; appointment scheduling; real-time communication; MERN stack; digital visitor pass; web application