Rail infrastructure plays an important role in freight and passenger mobility, and the assessment of rail track structure depends critically on understanding how the rail interacts with the supporting foundation. When rail support degrades (e.g., due to ballast fouling, settlement, etc.), the rail exhibits greater localized deformation that can lead to serious deleterious conditions. Track modulus represents a fundamental diagnostic measure of rail support, encompassing the vertical stiffness characteristics of the foundation and its resistance against downward rail movement. Existing track modulus characterization methodologies typically comprise deflection measurements of railway track (e.g., tie deflections) under known loads. Track modulus estimations result from analyzing deflection and load under assumptions of a traditional Winkler foundation, which can oversimplify mechanic relationships. Specifically, in the context of rail–ballast–subgrade interaction, a tensionless foundation permits gap development which can occur as track structure separates from the supporting ballast; additionally, track modulus may vary along the track length as conditions vary spatially. This paper presents a general analytical solution of ballasted track support characterization based on an iterative algorithm for the static response of a finite beam resting on a tensionless Winkler foundation. The method relates to multiple loads (e.g., concentrated axle loads and distributed self-weight), deflection along the track, and track condition through singularity functions, superposition of discrete support springs, and moment–curvature relationships. The model estimates rail deflections, lift-off points and shear and moment diagrams along the track. The technique permits: (1) validations against benchmark solutions and previously published results, (2) estimations of track modulus from known loads and measured deflections, and ultimately, (3) a framework for designing and processing sensor data streams for use in analyses and evaluations of railway track structure.

Laparoscopy is one of the most widely used surgical techniques in clinical practice. However, its practice is associated with medium- and long-term musculoskeletal disorders in surgeons. In this context, robot-assisted surgery has emerged as a promising approach for mitigating ergonomic constraints while enhancing control and precision during laparoscope manipulation. Despite these advances, existing research predominantly focuses on robotic control strategies, whereas the study of human–robot interaction in the operating room remains comparatively underexplored. This paper presents a proof-of-concept framework for workspace-aware posture adaptation in collaborative surgical robotics. The proposed approach combines vision-based human activity recognition with reinforcement learning to control the shoulder–elbow–wrist redundant angle of a seven-degree-of-freedom manipulator holding a laparoscope. Based on the detected interaction context, the system distinguishes between controlling, observing, cutting, and blocked states. During the observation and cutting phases, the controller allows the robot’s posture to be reconfigured so that it tilts away from the human operator while maintaining the position of the laparoscope; when the surgeon moves away, the robot gradually returns to its default configuration. Two reward formulations, dense and fuzzy, are compared. Real-world experiments show that both approaches learn the desired reflexive behavior, while the fuzzy reward yields improved training stability and more consistent real-system performance, increasing workspace availability around the surgeon.

Ground-based scatterometers are widely used for quantitative microwave backscattering measurements in soil moisture retrieval, vegetation monitoring, and satellite scatterometer validation. However, low-cost software-defined radio (SDR) transceivers provide limited instantaneous bandwidth, making it difficult to transmit and process signals with bandwidths on the order of hundreds of MHz for fine range resolution, especially for systems requiring real-time onboard processing. To address this problem, this paper presents a vehicular, fully polarimetric, SDR-based scatterometer that achieves an equivalent wideband response by sequentially transmitting adjacent narrow subbands and coherently synthesizing them onboard. To enable real-time operation on a resource-limited field-programmable gate array/system-on-chip (FPGA/SoC) platform, we adopt a frequency-domain synthesis-pulse-compression pipeline that avoids interpolation and eliminates repeated matched filtering across subbands. A slot-based online phase calibration is performed within the settling window after each fast lock to estimate and compensate random local oscillator (LO) phase offsets, preserving coherent stitching. In addition, pulse repetition within each subband and coherent accumulation are integrated to improve the signal-to-noise ratio (SNR) under real-time throughput constraints. A Zynq-based implementation demonstrates deterministic onboard range-profile output, with a minimum processing latency of about 1.57 ms per frame. Loopback and outdoor experiments validate the equivalent 200 MHz bandwidth (five 40 MHz subbands), achieving approximately 0.75 m resolution and yielding sidelobe metrics consistent with the designed windowing, including a peak sidelobe ratio (PSLR) of −27.43 dB and an integrated sidelobe ratio (ISLR) of −12.38 dB. Field scans over farmland further show consistent trends across incidence angle and azimuth, indicating reliable onboard quantitative backscattering measurement. These results demonstrate that the proposed method provides a feasible solution for deterministic real-time equivalent wideband scatterometry on a low-cost SDR platform.

Traditional deep learning-based models have achieved promising results in medical image segmentation. However, their performance degrades severely when applied to unseen domains due to variations in imaging protocols, acquisition devices, and patient populations across medical centers, which lead to significant distribution shifts. With the emergence of the Segment Anything Model (SAM), a single model now exhibits significantly improved generalization and adaptability to various image types. Nevertheless, while SAM has learned structure representations from large-scale natural images, it lacks fine-grained structural knowledge specific to the medical imaging domain, remaining relatively invariant across imaging domains. In addition, its structural enhancement is vulnerable to unreliable prompts, and patch-wise inference disrupts structural continuity, leading to suboptimal performance in capturing anatomical details. To address this, we propose a novel Medical Fine-grained Segment Anything Model (termed MedFineSAM), which integrates three key modules: shared fine-grained structural enhancement, which extracts and selectively enhances fine-grained structural features shared between prompts and image embeddings via a structural dictionary; a prompt gating mechanism, which estimates prompt confidence and dynamically adjusts prompt weights to avoid erroneous enhancement; and a structural continuity diffusion in frequency domain (SCFD), which performs frequency-domain smoothing during decoding to alleviate structural discontinuity caused by patch aggregation. Experiments on the fundus benchmark and prostate MRI benchmark demonstrate superior generalization performance, offering new insights into leveraging SAM for single-source domain generalization in medical image segmentation.

Distributed edge sensing systems, such as IoT monitoring nodes, wearable devices, and camera-based sensing terminals, continuously generate privacy-sensitive data that are costly to transmit to a central server. Federated learning (FL) provides a promising solution for collaborative model training without raw-data sharing; however, its practical deployment in edge sensing systems is challenged by non-IID local observations, limited uplink/downlink resources, and restricted on-device computation. To address these issues, this paper proposes a Dual-Sided Sparse Aggregation (DSSA) mechanism integrated with FedProx for resource-constrained edge sensing environments. In the proposed framework, the server prunes the global model after each communication round and transmits only the retained parameters, while clients update the complementary parameters and upload sparse local gradients. This fixed-structure sparse training strategy reduces bidirectional communication overhead and local computation cost, while FedProx improves robustness under heterogeneous data distributions. Experiments on CIFAR-10 and SVHN with varying non-IID degrees, pruning ratios, and hyperparameter settings show that the proposed method achieves a favorable resource-performance trade-off, reducing communication cost by up to 73.0% and computation cost by up to 34.9% while maintaining competitive accuracy. Under controlled benchmark settings, the proposed method demonstrates substantial resource savings compared with FedAvg, particularly in mildly heterogeneous scenarios, indicating a favorable benchmark-level resource-performance trade-off for resource-constrained edge sensing scenarios under the evaluated settings.

Infrared small target detection is widely used in aerospace surveillance, maritime search and rescue, and military reconnaissance. However, the performance of detection algorithms is highly dependent on scene characteristics, and methods that perform well in simple backgrounds may degrade substantially in complex environments. Existing indicators, such as information entropy, average gradient, and peak signal-to-noise ratio, can reflect detection difficulty from individual perspectives, but they do not provide a unified measure that jointly considers target saliency, background complexity, and target–background coupling. To address this issue, this study proposes a scene detection complexity (SDC) metric for quantifying the difficulty of infrared small target detection. Six basic indicators are selected from three dimensions, namely target saliency, background complexity, and target–background coupling: statistical variance, target–background contrast, signal-to-clutter ratio, information entropy, structural similarity, and target size. After Min–Max normalization, objective weights are determined by combining the entropy weight method and principal component analysis, and the weighted indicators are fused into an SDC value in the range of . Experiments on 100 test images selected from IRST640, MSISTD, SIRST-V2, and an infrared small-aircraft sequence dataset show that the proposed SDC achieves a Pearson linear correlation coefficient of 0.956 with subjective difficulty ratings and with image-level detection scores obtained from seven representative algorithms. The results further indicate that traditional methods are more sensitive to increasing scene complexity, whereas deep-learning-based methods are comparatively more robust in complex backgrounds. The proposed SDC provides a unified and objective tool for performance evaluation, algorithm selection, and pre-assessment of scene difficulty in infrared small target detection.

This Technical Report presents a quantitative signal processing approach to analyze and correct eye drift during vestibulo-ocular reflex (VOR) measurements using the video Head Impulse Test (vHIT). The objective is to determine the extent of drift caused by goggle slippage—a technical artifact that can distort the VOR gain index. A total of 57 impulses were categorized into three protocols: Lateral, LARP, and RALP. For each impulse, peak head velocity and eye drift (estimated from the average velocity during the pre- and post-impulse rest periods) were extracted using a custom signal processing pipeline implemented in MATLAB R2020b and Python 3.11 64 bit. Results showed the highest drift in the RALP group (−7.41 deg/s) and the lowest in the LARP group (−3.08 deg/s). The correlation between head velocity and drift was most prominent in the RALP group (r > 0.7), highlighting the impact of stimulation direction on goggle stability. This study proposes a drift detection method to be integrated into VOR correction algorithms, thereby enhancing gain analysis and saccade detection in automated systems.

Molecular-scale detection based on quantum tunnelling is promising for molecular electronics and high-sensitivity analysis, owing to its sensitivity to molecular structure and energy levels. However, conventional two-electrode tunnelling measurements suffer from overlapping conductivity of different molecules, limiting molecular discrimination in complex systems. To address this, we propose an electrochemical-gate-controlled nanoscale tunnelling strategy that expands the two-electrode system to a three-electrode configuration via a tunable gate potential, enabling the differentiation of distinct molecules at near-single-molecule sensitivity. Scanning the gate potential under constant tunnelling bias modulates the alignment between molecular orbitals and the electrode Fermi level, altering the statistical characteristics of molecular tunnelling transport. Experimental results show that target molecules induce a bimodal distribution of tunnelling current (background and molecule-correlated channels), with the second peak exhibiting distinct gate potential dependence. Comparative analysis of ascorbic acid (AA), acetylcholine (ACh), and uric acid (UA) reveals unique trajectories of characteristic peaks with gate potential, forming an electrochemical gate response fingerprint. This gate-dependent conductance trajectory provides a novel statistical dimension for molecular recognition, enabling differentiation of distinct molecules.

Video Anomaly Detection (VAD) is commonly formulated as a one-class classification task. Global motion, with temporal variations across most pixels within an object-centric region, e.g., walking, is typically regular, whereas localized motion, e.g., waving, can be ambiguous. Decoupled spatial and temporal jigsaw puzzles (DSTJiP) is a self-supervised method that learns discriminative representations by predicting the original order of spatially and temporally shuffled patches. However, DSTJiP’s uniform sampling and equal weighting do not assign stronger supervision to global-motion examples within the temporal objective. Consequently, the temporal supervision allocated to global-motion examples may become insufficient across training-data regimes with varying proportions of these examples, deteriorating VAD performance. Nevertheless, excessively strengthening such supervision also degrades performance. To address these issues, we propose spatial and extra temporal jigsaw puzzles (SETJiP) with two RGB-only training schemes that provide stronger and more conservative temporal supervision for global-motion examples, respectively. One scheme strengthens temporal supervision on these examples via additional temporal jigsaw puzzles. The other does so more conservatively by upweighting their temporal jigsaw puzzles. Experiments on four VAD benchmarks show that both schemes improve on DSTJiP and remain highly competitive with state-of-the-art methods.

Advanced sensor signal analysis is increasingly important for intelligent health management in human-centered environments, where continuous perception and real-time interpretation of motion-related signals are essential for safe and adaptive assistance. In this study, we propose a vision-based sensor signal analysis framework for automated ergonomic risk assessment of wheelchair users during cabinet interaction. The proposed framework integrates YOLOv11 for human detection, MHFormer for monocular 3D pose reconstruction, and a fuzzy logic-enhanced RULA model for continuous ergonomic risk quantification from video-derived motion signals. To support model development and evaluation, we constructed a dedicated wheelchair cabinet-operation dataset comprising 30 participants, including 14 experienced wheelchair users and 16 trained simulation participants, across five representative cabinet-operation scenarios. The raw dataset contained approximately 5 h of RGB video and about 150,000 original frames. To reduce redundancy caused by highly similar consecutive frames and to mitigate overfitting risk, representative frames were sampled from the continuous video sequences, resulting in 10,000 images for annotation and model development. Based on the proposed framework, raw visual sensor signals are transformed into temporally continuous kinematic representations and ergonomic risk scores, enabling non-contact and real-time health-state interpretation in assistive living environments. The proposed method achieved an average joint-angle estimation RMSE of 7.5°, representing an approximately 60% reduction compared with a Kinect v2-based motion capture baseline (18.6°), which is widely used for low-cost ergonomic evaluation. In benchmark evaluation, the proposed method achieved 84% risk-classification accuracy with a Cohen’s kappa of 0.66, outperforming representative baseline approaches. The results further indicated that low revolving-door and low-drawer operations were associated with higher and more sustained ergonomic risk exposure than sliding-door interaction. These findings demonstrate that vision-based sensor signal analysis can provide an effective solution for intelligent health management, ergonomic monitoring, and perception-driven assessment in accessible and assistive autonomous living systems.