Wearable eye-tracking technologies remain constrained by bulky optics, high power consumption, and reliance on external computation. We present a hardware-software codesigned electrooculography (EOG) interface that integrates ultrathin conformal e-skin sensors with resistive random-access memory (RRAM) crossbar, used to implement synaptic vector-matrix multiplication within a neuromorphic processing pipeline for real-time gaze decoding. Conformal e-skin sensors provide stable and continuous acquisition of both vertical and horizontal oculomotor signals, which are transformed into attention-guided spike features for classification by a lightweight spiking neural network (SNN). Implemented on an RRAM array with quantized weights after noise-aware training, the system achieves robust inference while substantially lowering latency and energy demand. The proposed framework is designed for local edge-level computation, thereby preserving user privacy and eliminating the need for cloud-based inference. By eliminating head-mounted optics, this glassless architecture enables unobtrusive and energy-efficient wearable interfaces. These results establish flexible bioelectronics with neuromorphic processors to advance immersive computing, assistive interfaces, and mobile health monitoring.

Epilepsy affects over 50 million people worldwide, yet automated seizure detection systems either achieve moderate sensitivity with excessive false alarms or rely on uninterpretable deep networks. This study presents a patient-independent EEG-based seizure detection framework that achieved zero false alarms in 24 h with 95% sensitivity in a retrospective evaluation on a CHB–MIT pediatric cohort (n = 6 seizure-positive patients). The pipeline extracts 27 time-, frequency-, and nonlinear-domain features from 5 s windows and trains five ensemble classifiers (XGBoost, CatBoost, LightGBM, Extra Trees, Random Forest) using strict leave-one-subject-out cross-validation. All models achieved segment-level AUC ≥ 0.99. Under zero-false-alarm constraints, XGBoost attained perfect specificity with 0.922 sensitivity. SHAP and LIME analyses suggested candidate EEG biomarkers that appear consistent with known ictal signatures, including temporo-parietal theta-band power, amplitude variability (IQR, RMS), and Hjorth activity. External validation on the Siena Scalp EEG Database (12 adult patients, 37 seizures) demonstrated cross-dataset generalization with 95% event-level sensitivity (Extra Trees) and AUC of 0.86 (Random Forest). Temporal lobe channels dominated feature importance in both datasets, confirming consistent biomarker identification across pediatric and adult populations. These findings demonstrate that calibrated gradient-boosted ensembles using interpretable EEG features achieve clinically safe seizure detection with cross-dataset generalizability.

Most authentication models are vulnerable to security breaches when personal data is exposed. This study introduces a novel hybrid visual computer interface integrating event-related potentials (ERPs) and steady-state visually evoked potentials (SSVEPs) to develop an authentication system that enhances both performance and personalization in neural interfaces. Our model utilizes distinctive neural patterns elicited by a range of visual stimuli based on 4-digit numbers, such as familiar numbers (personal birthdates, excluding targets), standard targets, and non-targets. The results revealed a distinct P300 response to familiar numbers when compared to both non-target and target stimuli. Incorporating these stimuli into our Transformer-based authentication system, coupled with personalized electroencephalogram (EEG) data segmentation, resulted in high accuracy in authenticating users and demonstrated remarkable robustness against security breaches. Additionally, a 10 Hz grow/shrink background image successfully elicited SSVEP. Furthermore, the comparison of harmonic and fundamental frequencies aids in optimizing neural interfaces.

The growing demand for secure, immersive authentication in extended reality (XR) environments calls for neural interfaces that are both robust and user-friendly. This study introduces a novel and robust dual-modality EEG-based authentication framework that independently exploits: 1) steady-state visually evoked potentials (SSVEP) and 2) eye-blink-induced EEG responses as covert neural signatures. Both signals are recorded using a 64-channel EEG system seamlessly integrated with the Microsoft HoloLens 2 for immersive XR-based user evaluation. To mitigate visual fatigue while preserving signal fidelity, we replace conventional flicker stimuli with a 10 Hz grow-shrink visual design. We employ a modality-specific classification strategy, modeling SSVEP and eye-blink signals independently to retain their distinct neurophysiological characteristics. A multi-stage feature selection pipeline combines SHAP and Random Forest rankings, followed by logistic regression-based permutation importance to identify the top 10 discriminative features per modality. These features undergo statistical validation via non-parametric tests to ensure physiological plausibility and class separability. Classification is subsequently performed using four machine learning models-Random Forest, XGBoost, Support Vector Machine, and Logistic Regression-with Random Forest and XGBoost consistently yielding the highest performance. Evaluated across 20 participants using user-wise validation, our framework achieves over 99% accuracy and near-perfect ROC-AUC scores for both modalities, confirming strong discriminability between genuine and impostor attempts. Our results demonstrate that interpretable, fatigue-aware EEG features can deliver high authentication performance under XR conditions. The proposed system is lightweight and explicitly engineered for real-time deployment and spoof-resistance, making it well-suited for future XR-based defense, training, and industrial applications.

Neuromorphic systems are emerging as a promising alternative to revolutionize silicon-based computing devices. Two-dimensional materials are considered promising candidates for active materials due to their unique advantages. This review provides a comprehensive overview of neuromorphic computing based on 2D materials, encompassing the development from the synaptic devices based on 2D materials to the neuromorphic system demonstration. Besides, we further outline various applications of neuromorphic computing in the emerging applications of neuromorphic computing

Photodiodes based on two-dimensional semiconductors are of potential use in the development of optoelectronic devices, but their photovoltaic efficiency is limited by strong Fermi level pinning at metal–semiconductor contacts. Typical metal–interlayer–semiconductor contacts can address this issue, but can also lead to an increase in series resistance. Here we report a conductive-bridge interlayer contact that offers both Fermi level depinning and low resistance. We create an oxide interlayer that decouples the metal and semiconductor, while embedded gold nanoclusters in the interlayer act as conductive paths that facilitate efficient charge transport. Using these contacts, we fabricate a tungsten disulfide (WS2) photodiode with a photoresponsivity of 0.29 A W−1, linear dynamic range of 122 dB and power conversion efficiency of 9.9

This study introduces a machine learning–driven extended reality (XR) interaction framework that leverages electroencephalography (EEG) for decoding consumer intentions in immersive decision-making tasks, demonstrated through functional food purchasing within a simulated autonomous vehicle setting. Recognizing inherent limitations in traditional “Preference vs. Non-Preference” EEG paradigms for immersive product evaluation, we propose a novel and robust “Rest vs. Intention” classification approach that significantly enhances cognitive signal contrast and improves interpretability. Eight healthy adults participated in immersive XR product evaluations within a simulated autonomous driving environment using the Microsoft HoloLens 2 headset (Microsoft Corp., Redmond, WA, USA). Participants assessed 3D-rendered multivitamin supplements systematically varied in intrinsic (ingredient, origin) and extrinsic (color, formulation) attributes. Event-related potentials (ERPs) were extracted from 64-channel EEG recordings, specifically targeting five neurocognitive components: N1 (perceptual attention), P2 (stimulus salience), N2 (conflict monitoring), P3 (decision evaluation), and LPP (motivational relevance). Four ensemble classifiers (Extra Trees, LightGBM, Random Forest, XGBoost) were trained to discriminate cognitive states under both paradigms. The ‘Rest vs. Intention’ approach achieved high cross-validated classification accuracy (up to 97.3% in this sample), and area under the curve (AUC > 0.97) SHAP-based interpretability identified dominant contributions from the N1, P2, and N2 components, aligning with neurophysiological processes of attentional allocation and cognitive control. These findings provide preliminary evidence of the viability of ERP-based intention decoding within a simulated autonomous-vehicle setting. Our framework serves as an exploratory proof-of-concept foundation for future development of real-time, BCI-enabled in-transit commerce systems, while underscoring the need for larger-scale validation in authentic AV environments and raising important considerations for ethics and privacy in neuromarketing applications.

Pupillary-computer interface (PCI) refers to a novel interaction modality that leverages pupil size variations elicited by changes in visual stimulus brightness. The PCI based on the pupillary light reflex (PLR) induced by binary-coded visual stimuli was proposed. A novel PCI interface was devised to overcome the limitations of conventional electroencephalogram hardware, using artificial intelligence to model subtle pupil signal patterns induced by visual stimuli. The proposed PCI system exhibited high performance in terms of the number of commands, classification accuracy, and information transfer rate (ITR) using a simple binary coding scheme and convolutional neural network-based deep learning. Twelve healthy subjects (six men and six women, aged 28.6 ± 3.4 year) participated in three experimental conditions, each using 4-, 10-, and 20-class binary-coded visual stimuli. Each visual stimulus was constructed by dividing the 3-s period into ten phases of 0.3 s each, with a single brightness change (e.g., from dark to bright) occurring within this interval. The proposed system achieved a high classification accuracy (91.84 %, 93.84 %, and 98.61 %) and ITR (59.74, 62.04, and 69.36 bits/min) for 20-, 10-, and 4-class stimuli in the test dataset, considerably outperforming previous PLR-based interface studies. The findings indicate that the proposed PCI system provides a simple, cost-effective, and low-training-requirement interface solution that does not require user training and maintains long-term stability.

Solid oxide fuel cells (SOFCs) are all-solid-state electrochemical devices that directly convert the chemical energy of fuel into electricity, and they represent one of the most efficient and versatile means of clean power generation. Herein, we report the potential of SOFCs as neuromorphic computing devices and verify the synaptic properties, including paired-pulse facilitation (PPF) and potentiation and depression characteristics, under specific operating conditions, which enable the implementation of a brain-like computing system. We simulate image classification tasks using an artificial neural network (ANN) and confirm that the classification accuracy is 85.4%, indicating that the SOFC system is capable of performing cognitive computing in addition to having power generation functionality. This work examines the concept of in-cell computing using bifunctional SOFCs that handle computation/memory functions in electric power generation devices, and our findings suggest opportunities for power management in various sectors, including smart grid systems, electric vehicles, and smart mobility.

The purpose of this study was to propose a model for estimating individual susceptibility to motion sickness by correlating measures of seated postural sway before exposure to a roller coaster movie displayed in a head-mounted display (HMD) with subjective ratings of visually induced motion sickness (VIMS). The participants were required to watch the content for 15 min, and seated center of pressure (sCOP) was measured using a force platform for 5 min before viewing. We developed a VIMS estimating model from 16 subjects and verified it on 15 subjects. SSQ scores for VIMS were strongly correlated with the area of sCOP, the main-to-minor ratio of the sCOP ellipse, and the total length of sCOP, respectively (r = 0.706, r = 0.555, and r = 0.622). When verifying the estimation model, the observed SSQ scores differed by 32.628 +/- 29.749 from the expected SSQ scores, and both score sets were positively correlated (r = 0.622). The classification results between nil and mild, moderate, and severe groups for VIMS showed an accuracy of 0.74 using random under-sampling boost (RUSBoost).

The operational efficacy of lane departure warning systems (LDWS) in autonomous vehicles is critically influenced by the retro-reflectivity of road markings, which varies with environmental wear and weather conditions. This study investigated how changes in road marking retro-reflectivity, due to factors such as weather and physical wear, impact the performance of LDWS. The study was conducted at the Yeoncheon SOC Demonstration Research Center, where various weather scenarios, including rainfall and transitions between day and night lighting, were simulated. We applied controlled wear to white, yellow, and blue road markings and measured their retro-reflectivity at multiple stages of degradation. Our methods included rigorous testing of the LDWS’s recognition rates under these diverse environmental conditions. Our results showed that higher retro-reflectivity levels significantly improve the detection capability of LDWS, particularly in adverse weather conditions. Additionally, the study led to the development of a simulation framework for analyzing the cost-effectiveness of road marking maintenance strategies. This framework aims to align maintenance costs with the safety requirements of autonomous vehicles. The findings highlight the need for revising current road marking guidelines to accommodate the advanced sensor-based needs of autonomous driving systems. By enhancing retro-reflectivity standards, the study suggests a path towards optimizing road safety in the age of autonomous vehicles.

IntroductionThe rising prevalence of obesity has become a public health concern, requiring efficient and comprehensive prevention strategies.MethodsThis study innovatively investigated the combined influence of individual and social/environmental factors on obesity within the urban landscape of Seoul, by employing advanced machine learning approaches. We collected ‘Community Health Surveys’ and credit card usage data to represent individual factors. In parallel, we utilized ‘Seoul Open Data’ to encapsulate social/environmental factors contributing to obesity. A Random Forest model was used to predict obesity based on individual factors. The model was further subjected to Shapley Additive Explanations (SHAP) algorithms to determine each factor’s relative importance in obesity prediction. For social/environmental factors, we used the Geographically Weighted Least Absolute Shrinkage and Selection Operator (GWLASSO) to calculate the regression coefficients.ResultsThe Random Forest model predicted obesity with an accuracy of >90%. The SHAP revealed diverse influential individual obesity-related factors in each Gu district, although ‘self-awareness of obesity’, ‘weight control experience’, and ‘high blood pressure experience’ were among the top five influential factors across all Gu districts. The GWLASSO indicated variations in regression coefficients between social/environmental factors across different districts.ConclusionOur findings provide valuable insights for designing targeted obesity prevention programs that integrate different individual and social/environmental factors within the context of urban design, even within the same city. This study enhances the efficient development and application of explainable machine learning in devising urban health strategies. We recommend that each autonomous district consider these differential influential factors in designing their budget plans to tackle obesity effectively.

Three-dimensional (3D) hetero-integration technology is poised to revolutionize the field of electronics by stacking functional layers vertically, thereby creating novel 3D circuity architectures with high integration density and unparalleled multifunctionality. However, the conventional 3D integration technique involves complex wafer processing and intricate interlayer wiring. Here we demonstrate monolithic 3D integration of two-dimensional, material-based artificial intelligence (AI)-processing hardware with ultimate integrability and multifunctionality. A total of six layers of transistor and memristor arrays were vertically integrated into a 3D nanosystem to perform AI tasks, by peeling and stacking of AI processing layers made from bottom-up synthesized two-dimensional materials. This fully monolithic-3D-integrated AI system substantially reduces processing time, voltage drops, latency and footprint due to its densely packed AI processing layers with dense interlayer connectivity. The successful demonstration of this monolithic-3D-integrated AI system will not only provide a material-level solution for hetero-integration of electronics, but also pave the way for unprecedented multifunctional computing hardware with ultimate parallelism.

Objective: The purpose of this study was to investigate whether electromyogram (EMG) induced by jaw-clenching can be accurately observed in steady-state visually evoked potential (SSVEP) signals at occipital lobes as well as temporal lobes. We also investigated a feasibility that the fused SSVEP data with the jaw-clenching EMG can be classified as a function of task types characterized by SSVEPs with and without the short and long clenching noise. Based on this hypothesis, we proposed a novel command protocol to effectively and efficiently reduce false alarms caused by using SSVEP brain-computer interface (BCI).BRBRBackground: Much attention has been paid to various brain-computer interfaces for neural rehabilitation and alternative communication channels for disabled people with severe neurological disorders. However, despite the interests in BCI, false alarms caused by interference effects between command flickers with different frequencies or internal and external noises have been still problematic and have yet to be investigated. Even though protocols to correct the false alarms based on error-related potential (ErrP) or eye blinks have been proposed, it has still limitation due to errors inherent in involuntary reactions of humans. Thus, it is necessary to develop a protocol to address the issue and propose a method insusceptible to the involuntary reactions.BRBRMethod: Nine undergraduate students (3 female) voluntarily participated in the experiment. They were asked to divide their attention between the two squares randomly oscillating with 8.57 and 10Hz and perform three types of tasks characterized by SSVEP only, short clenching, and long clenching conditions. They carried out the three tasks with a semi-counterbalanced and random order and rest interval between the tasks was set to 10 minutes to minimize order and carryover effects. It took about 12 minutes in completing the experimental tasks for each participant. We used five machine learning algorithms of binary classification to compare the performance in classifying the brainwave signals characterized by the three tasks.BRBRResults: As a result of the classification comparison, random forest showed the highest classification performance of more than 0.85 (AUC) as follows: 0.88 (short clenching vs. SSVEP), 0.87 (long clenching vs. SSVEP), 0.85 (long clenching vs. short clenching condition) in the test data set. Notably, the findings implicated a clear classification in the long cand short clenching condition even at the frequency band of more than 20Hz. For the participants whose highest weight factor was shown at the occipital lobe, it can be possible to reduce false alarms in the SSVEP BCI with only single electrode at the occipital lobe.BRBRConclusion: The findings found in this study implicate that the proposed protocol can be used to effectively and efficiently correct false alarms in event-related potential (ERP), event-related desynchronization/synchronization (ERD/ERS) BCI as well as SSVEP BCI. On top of that, classifying the subtle changes in the jaw clenching types enables BCI developers to create additional commands to modulate their BCI system. The results can be greatly improved if weight factors in the binary classification are modulated as a function of individual characteristics.BRBRApplication: The findings obtained in this study can be utilized in developing asynchronous BCI very robust to false alarms and expected to provide a future direction into hybrid BCI system.

The event-related potential (ERP) technique is widely used in various fields, but the requirement for sensor attachment limits its application. The aim of this study was to develop an infrared (IR) webcam-based, non-contact system to obtain the amplitudes and latencies of ERP via measurement of event-related pupillary responses (ErPRs) from pupillary rhythms. A total of 32 healthy volunteers participated in this study, and they were required to perform the three levels of mental arithmetic tasks to induce mental workloads (MWLs). ERPs and ErPR were measured by ERP task for the different MWL states and compared based on statistical analysis, classification, correlation, and Bland-Altman plot. Both ERP and ErPR amplitudes and latencies for the three MWL states were revealed by one-way repeated measures analysis of variance to be statistically significantly different. Furthermore, the ERP and ErPR data were found to have 73 and 80% classification performances, respectively, using k-nearest neighbour (10-fold cross validation, n = 96). Correlation coefficients between ERP and ErPR features, which were in the range of 0.642-0.734, indicated good or strong correlation, and good agreement between the indices from these two types of experimental measurement indices was apparent in Bland-Altman plots. An advanced method for IR-webcam-based non-contact determination of ERP components via ErPR measurement was successfully developed, and it was demonstrated that this technique can be used for ERP component monitoring using a low-cost, non-invasive, and easily implemented IR webcam without the burden of sensor attachment.

This study investigated the effects of modulated respiration on blood pressure and autonomic balance to develop a healthcare application system for stabilizing autonomic balance. Thirty-two participants were asked to perform self-regulated tasks with 18 different respiration sequences, and their electrocardiograms (ECG) and blood pressure were measured. Changes in cardiovascular system functions and blood pressure were compared between free-breathing and various respiration conditions. Systolic and diastolic blood pressures stabilized after individual harmonic breathing. Autonomic balance, characterized by heart rate variability, was also stabilized with brief respiration training according to harmonic frequency. Five machine-learning algorithms were used to classify the two opposing factors between the free and modulated breathing conditions. The random forest models outperformed the other classifiers in the training data of systolic blood pressure and heart rate variability. The mean areas under the curves (AUCs) were 0.82 for systolic blood pressure and 0.98 for heart rate variability. Our findings lend support that blood pressure and autonomic balance were improved by temporary harmonic frequency respiration. This study provides a self-regulated respiration system that can control and help stabilize blood pressure and autonomic balance, which would help reduce mental stress and enhance human task performance in various fields.

Both physiological and neurological mechanisms are reflected in pupillary rhythms via neural pathways between the brain and pupil nerves. This study aims to interpret the phenomenon of motion sickness such as fatigue, anxiety, nausea and disorientation using these mechanisms and to develop an advanced non-contact measurement method from an infrared webcam. Twenty-four volunteers (12 females) experienced virtual reality content through both two-dimensional and head-mounted device interpretations. An irregular pattern of the pupillary rhythms, demonstrated by an increasing mean and standard deviation of pupil diameter and decreasing pupillary rhythm coherence ratio, was revealed after the participants experienced motion sickness. The motion sickness was induced while watching the head-mounted device as compared to the two-dimensional virtual reality, with the motion sickness strongly related to the visual information processing load. In addition, the proposed method was verified using a new experimental dataset for 23 participants (11 females), with a classification performance of 89.6% (n = 48) and 80.4% (n = 46) for training and test sets using a support vector machine with a radial basis function kernel, respectively. The proposed method was proven to be capable of quantitatively measuring and monitoring motion sickness in real-time in a simple, economical and contactless manner using an infrared camera.