Daegu Gyeongbuk Institute of Science and Technology

Facial expressions are powerful signals of human emotion, shaping both human–human and human–computer interaction. As interactive technologies, from adaptive interfaces to emotion-aware agents, become more pervasive, systems are increasingly expected to recognize and respond to users’ emotions naturally. But what if a system misreads your face? Such misinterpretation is particularly likely when cultural differences in emotion perception are overlooked. This problem may be compounded by the fact that most facial emotion recognition (FER) models are trained on datasets that reflect the norms of a particular cultural group that assume universality, limiting their reliability in multicultural contexts. Surprise, in particular, is an emotion whose valence can be either positive or negative depending on context, making it a critical case for investigating cultural bias in FER. To address this, we examined how cultural background shapes the recognition and valence interpretation of surprise facial expressions among South Korean (N=36) and American (N=34) participants. Participants labeled 200 facial expressions (surprise and fear), rated their perceived valence, and described personal experiences of surprise. Results show that South Korean-labeled surprise expressions exhibited stronger negative Action Unit (AU) activation and lower valence ratings, whereas American-labeled ones showed more balanced or positive facial cues. Qualitative accounts further revealed that South Koreans framed surprise as tense or socially cautious, while Americans viewed it as open and situationally flexible. These findings bridge recognition and interpretation in cross-cultural emotion research and highlight the need for culturally adaptive FER systems that can interpret ambiguous emotions like surprise more inclusively.

As multifunctional chemical tools, organodiboron compounds present an important challenge in organic synthesis, with respect to their synthesis and functionalization. Although readily available 1,3-enynes have been employed as a platform for various regioselective difunctionalization reactions, the diborylation reactions of 1,3-enynes remain limited, and the installation of a CF3 group is often a prerequisite. In this study, we report a copper-catalyzed selective diborylation reaction of 1,3-enynes to access synthetically useful 1,1- and 1,4-diborylalkenes. The synthetic utility of this method is demonstrated by a gram-scale synthesis of a natural antifouling agent. Furthermore, the Cu-catalyzed trans-protoborylation reaction of aryl-substituted (Z)-enynes is reported. The thorough computational studies and the deuterium-labeling experiments provide insights into the reaction mechanism and the regio- and stereoselectivity of diborylated products.

Achieving both safe contact transitions and high-performance motion tracking is essential for robots manipulating in contact-rich environments. Disturbance-observer (DOB) frameworks enable robust motion control, but achieving compliant interaction typically requires additional strategies, and prior work has paid limited attention to how acceleration estimation affects in DOB behavior. In this work, we adopt a second-order sliding-mode observer (SOSML) to obtain noise-robust, low-lag acceleration estimates and integrates them into an Impedance-Aware Disturbance Observer (IADOB). The proposed controller integrates DOB feedback with inner force control to exploit the robot’s natural inertia and rendered stiffness, enabling safe contact transitions without sacrificing motion tracking performance. Experiments on a 7-degree-of-freedom (DoF) Franka Research 3 robot demonstrate improved contact stability, and higher admissible stiffness compared to conventional impedance and momentum observer-based controllers.

Magnetic manipulation is increasingly used in medical applications for its potential in remote control. However, precise magnetic field generation in large workspaces remains challenging. This paper introduces an adaptive robotic end effector, the tunable magnetic end effector (TME), capable of generating spatially controllable magnetic fields. By integrating permanent magnets, the TME enables accurate magnetic control for wireless manipulation of miniaturized medical devices. Compared to standard switchable permanent magnets, TME offers enhanced field control suited for delicate operations. Finite element (FEM) simulations and experiments confirm reliable ON/OFF field switching, showing a 7.2% average error. Key design parameters (magnet size, material, and arrangement) were optimized via simulation. An artificial neural network (ANN), trained on spatial, rotational, and magnetic data, enables adaptive control. Proof-of-concept demos include steering millimeter-scale magnetic carriers, shaping magnetic soft robots, and directing magnetic nanoparticle swarms. The dual-TME configuration further expands the effective manipulation workspace and enables dynamic switching of magnetic field directions across different regions, thereby enhancing the system's applicability.

This paper presents a unified deep learning framework for multi-class detection and tracking of visually similar road users in complex road environments. The proposed system is designed to accurately distinguish and track various road entities—such as cars, buses, trucks, bicycles, motorcycles, pedestrians, and riders—which often exhibit similar visual features. To achieve robust detection and instance-aware tracking, we adopt a three-stage training strategy: (1) a supervised multi-task learning stage for joint object detection and class identification, (2) a self-supervised contrastive learning stage for extracting instance-level feature embeddings without identity labels, and (3) a fine-tuning stage to improve object identification accuracy by refining both the detection and feature embedding heads. The unified network simultaneously outputs object classes, locations, and appearance embeddings, enabling consistent identity association across video frames. Experimental results demonstrate that the proposed method outperforms existing approaches in both fine-grained detection and multi-object tracking accuracy.

Seborrheic dermatitis and psoriasis are skin conditions that can adversely affect patients’ mental health and quality of life. Despite overlapping early-stage features, they differ in cause and treatment, making early and accurate diagnosis essential. To address this, we present a deep learning-based diagnostic solution combining a mobile multimodal imaging system with a Transformer-based Iterative Fusion Classification Model. This platform non-invasively captures multispectral, polarization, and fluorescence images of skin lesions, offering high-contrast views of tissue structure and composition. Unlike traditional CNNs, our Transformer architecture iteratively fuses data across modalities using attention mechanisms, enabling it to detect subtle differences between the two diseases. This leads to more accurate and robust classification. Clinical validation shows that our model outperforms conventional methods, demonstrating the system’s potential for accessible and precise mobile skin disease screening, ultimately supporting earlier intervention and improved patient outcomes.

Sweat monitoring offers real‐time insights into physiological conditions such as hydration, muscle fatigue, and metabolic status. However, conventional sweat sensors often face challenges associated with unstable skin contact and insufficient sampling. In this study, a fiber‐based wearable sensing platform is proposed, which incorporates semiconducting molybdenum disulfide (MoS 2 ) and polylactic acid (PLA) composite fibers fabricated via wet spinning. By exploiting the high surface‐to‐volume ratio and n‐type semiconducting nature of the MoS 2 network, the sensor selectively detects major biomarkers including electrolytes (Na + and K + ) and metabolites (lactic acid and NH 4 + ) via distinct electrostatic screening and charge trapping mechanisms. Furthermore, the intrinsic capillary action and thermal insulation of the fibers ensured reliable sweat collection without the requirement for external power. Additionally, the composite fiber exhibits piezoresistive capabilities, enabling simultaneous pressure monitoring to track physical motion. Multifunctional sensing facilitates the early diagnosis of metabolic disorders and the precise tracking of athletic performance. The developed fiber‐based sensor provides a robust textile‐integrated solution for next‐generation personalized healthcare monitoring.