Conference Agenda

Electroencephalogram (EEG) recordings are highly susceptible to ocular artifacts such as eye blinks and eye movements, which degrade signal quality. In addition, it affects the subsequent neurophysiological analysis and brain-computer interface (BCI) performance. Although recent deep learning based denoising approaches have shown promising results, most of them still lack selective global refinement and tend to over smooth EEG waveforms when applied directly to signal reconstruction. This paper proposes an efficient framework that integrates multiscale convolutional feature extraction, bidirectional long short-term memory (BiLSTM) based temporal modelling and residual kernel attention based artifact correction. Unlike conventional attention based denoising models, kernel attention is reformulated as a constrained residual refinement mechanism operating on top of a BiLSTM based reconstruction backbone. This enables the selective suppression of ocular artifacts while preserving neural signal quality. The proposed model has been evaluated on the EEGdenoiseNet open-source dataset for ocular artifact removal and compared against baseline models such as denoising autoencoders, 1D U-Net, multi-scale CNN and BiLSTM. Experimental results demonstrate that the proposed model achieves the highest average ΔSNR of 11.99 dB, along with lowest average RRMSE of 0.3813 and the highest correlation coefficient of 0.9245, indicating superior artifact suppression and waveform preservation. These results confirm the effectiveness and robustness of the proposed framework for practical EEG denoising applications.

Flexible and stretchable electronics are moving from laboratory prototypes to deployable products, particularly in passive ultra-high-frequency (UHF) RFID platforms for the Internet of Things (IoT) and wearable health technology. A persistent bottleneck is reliability under deformation: thin metal conductors crack, and conventional interconnects and adhesive joints often fail at the interfaces first. Self-healing conductors provide a materials-level route to improve robustness by restoring electrical continuity, but practical impact depends on packaging choices, especially chip attachment, interlayer bonding, and manufacturable patterning methods compatible with soft substrates. This work demonstrates fully self-healing UHF RFID tags featuring direct attachment of the IC to radiator ensuring robust interfaces under deformation. The radiators are fabricated from a soft self-healing supramolecular elastomer composite that combines electrical conductivity, strain-insensitivity, and pressure-sensitive adhesive properties. The material architecture embeds 200 nm-sized liquid-metal microparticles and functionalized carbon nanofillers within a dynamic polymer network, enabling a dynamically rearranging percolated conductive pathways that support autonomous interfacial bonding and reconnection. The conductor is solution-processable and can be additive manufactured into any tag antenna geometries, providing precise feature definition without the conventional deposition or subtractive patterning steps typical of conventional metal fabrication, such as wet etching. In the mechanical stack, the self-healing conductor is combined with a CNT-modified self-healing elastomer substrate, which enhances toughness and environmental stability, and improves interlayer bonding through pressure-sensitive adhesion. A central packaging element is the removal of a separate adhesive layer at the chip interface, improving interfacial robustness. RFID ICs mounted on a small rigid carrier with copper pads are attached directly onto the self-healing conductor by leveraging the material’s interfacial affinity and adhesive properties, forming an electrically stable and adaptable contact while avoiding an additional glue interphase that can delaminate or crack. The approach simplifies assembly and supports additive patterning using low-temperature, scalable printed electronics methods. The rigid packaged IC remains the least compliant component in the stack, motivating continued development of compliant pad layouts, local strain-relief features, and thin encapsulation strategies for highly deformable use cases. Overall, the work links conductive composite engineering with practical interface design, positioning self-healing UHF RFID as a packaging-driven route toward long-lived, deformable IoT and health technology devices.

Chronic wounds such as venous leg and diabetic foot ulcers are primarily managed through experienced clinical assessment; however, continuous, quantitative monitoring could support timely interventions, improving patient outcomes and potentially reducing nursing workload.

We report a skin-conformal smart wound dressing built on a bio-based thermoplastic polyurethane (TPU) substrate that integrates temperature and pH sensing within a flexible patch. The patch uses conventional PCB-based copper (Cu) trace technology, adapted through subtractive microfabrication to meet the demands of ultra-thin, stretchable TPU. The thinnest patch measures about 50 μm total thickness (two layers of 25 μm TPU). A reusable readout unit, provided by a partner, handles data acquisition, processing, and wireless transmission to a cloud platform, enabling remote clinical oversight while the disposable patch remains in contact with the wound bed. The pH sensor and reference electrode are supplied by a partner team. Ethylene oxide (ETO) sterilization was performed on assembled patches and demonstrated compatibility with the TPU substrate.

Mechanical characterization included cyclic stretch tests to simulate patient movement and dressing flexion. The bio-based TPU maintained elasticity and preserved interconnect integrity. Changes in local temperature and exudate pH reflect the wound’s inflammatory response and often indicate the need for a dressing change. In wounds with low infection suspicion, prolonged dressing intervals help preserve newly formed tissue, reduce pain, and avoid unnecessary workload and waste.

End-of-life strategy separates a multi-cycle reusable readout unit from a disposable patch hosting sensors with a lifetime of 10–14 days. The modular design minimizes waste while maintaining safety and clinical utility. Ongoing tests using a phantom leg model with mimicking a chronic wound will give usability and long-term results. Industrial fabrication of larger numbers of TPU based flexible and stretchable patches is under evaluation to meet high and stable quality standards.

Compact and portable self-charging power sources (SCPS) are envisioned as sustainable and comprehensive solutions to mitigating battery waste and promoting renewable energy utilization for distributed and wearable electronics. Higher-level sustainability and wider applicability can be anticipated if the entire SCPS are monolithically fabricated on paper substrates with small form factors and minimized use of electronic components and noble metals. Triboelectric generators, which harvest mechanical energy through vibration have attracted significant attention due to their abundant low-cost material choices, lightweight nature, and ability to be easily integrated onto paper. However, the pulsed, high-voltage, low-current output of triboelectric generators requires dedicated power management modules, which can hardly be fully integrated on paper substrates. Micro-supercapacitors (MSC), which have the advantages of excellent rate capability, high power density, and facile on-paper integration are well-suited energy storage devices for on-paper SCPS. Here, we report a scalable technique for monolithic fabrication of high-efficiency on-paper SCPS based on triboelectric generator and switched large-scale MSC arrays. Through direct ink writing of two types of carbon-based conductive inks and one type of gel electrolyte, combined with cutting, bending, and bonding of the paper substrate, a compact on-paper SCPS is integrated within a small footprint area of around 5 cm × 8 cm. Utilizing the strong electronegativity of PTFE film and the flexibility of paper, the triboelectric generator can deliver open-circuit peak voltage up to 200 V at 10 Hz with no evident performance degradation even after 100,000 contact-separation cycles. MSC arrays connected in series, with a large working voltage window of over 200 V and a capacitance of around 1 µF, can be charged to over 100 V within minutes by the high-voltage pulsed input. Within the on-paper SCPS, by folding the paper substrate, the 100 cells MSC array can be switched from series to parallel, achieving a voltage conversion ratio of 10:1. The resulting energy storage efficiency of this on-paper SCPS reaches a maximum of over 30%, comparable to the high efficiencies previously reported for general SCPS using optimized power management modules. We also demonstrated an on-paper SCPS based on a small scale MSC array without series to parallel switching, resulting in an effective power of 4 µW, sufficient to power an integrated commercial 3.3 V LED for 8 seconds after 10 minutes of 10 Hz mechanical flapping.

Full‑autonomous‑drive (FAD) vehicles rely on a heterogeneous suite of sensors—including lidar, radar, and vision systems—to perceive road environments and monitor surrounding pedestrians and traffic participants. While these sensing modalities perform reliably in open‑road scenarios, their effectiveness degrades significantly in dense parking environments where vehicles are tightly packed and line‑of‑sight is obstructed. In such conditions, autonomous vehicles struggle to achieve mutual identification, leading to ambiguity in vehicle‑to‑vehicle (V2V) and vehicle‑to‑everything (V2X) coordination. This limitation directly impacts localization accuracy, situational awareness, and the convergence speed of simultaneous localization and mapping (SLAM) algorithms.

To address this challenge, this paper presents a battery‑less, flexible ultrasonic radar‑reflector tag designed specifically for passive identification among autonomous vehicles in constrained environments. The proposed tag operates without an onboard power source; instead, it is acoustically driven by the host vehicle’s existing audio system. The resulting mechanical vibrations are captured remotely using a 120‑GHz millimeter‑wave radar, which demodulates the vibration signature through a unique equivalent‑sampling method. By encoding a lightweight digital identifier into the vibration pattern, each tag transmits a unique ID code that can be reliably extracted by nearby vehicles equipped with mmWave radar.

When integrated into a V2X network, these passive tags provide a robust mechanism for mutual identification, enabling autonomous vehicles to recognize neighboring nodes even when optical or lidar‑based line‑of‑sight is blocked. This capability is particularly valuable in dense parking lots, multi‑level garages, and urban curbside environments where conventional sensing is impaired. Experimental results demonstrate that the tag‑radar system achieves stable ID recovery at practical distances and angles, with minimal interference from ambient noise or multipath reflections. Furthermore, incorporating tag‑based identification into the SLAM pipeline significantly reduces iteration counts and improves localization stability, especially in feature‑poor or visually occluded spaces.

Overall, the proposed battery‑less ultrasonic radar‑reflector tag represents a low‑cost, maintenance‑free, and infrastructure‑independent solution for enhancing autonomous vehicle coordination in dense environments. By enabling passive, radar‑readable identification, the system strengthens V2X communication, improves SLAM performance, and provides a scalable pathway toward safer and more reliable full‑autonomous‑drive operation in real‑world deployment scenarios.