Search Results (855)

Electric bicycles are popular due to their environmental benefits and convenience. However, electric bicycle fires caused by series arc faults remain a serious safety concern. This study focuses on series arc fault identification for electric bicycles under complex operating conditions, covering state of charge (SoC), torque, and speed variations, and simultaneously considers normal state, DC-side series arc fault, and AC-side series arc fault conditions. Five time-domain features, namely root mean square (RMS), standard deviation (STD), skewness (SK), kurtosis (KUR), and current amplitude (CA), and three frequency-domain features, namely amplitude–frequency energy (AFE), amplitude–frequency mean (AFM), and amplitude–frequency kurtosis (AFK), are extracted. An improved principal component analysis (PCA)-based feature fusion method transforms the eight original time–frequency features into a five-dimensional PCA-fused feature representation consisting of PC1, PC2, PC3, fused PC4–PC7, and PC8. The fused features are classified using a radial basis function (RBF)-support vector machine (SVM) model. The proposed method achieves 98.68% test accuracy, 0.9869 Macro-F1, and 0.9931 Macro-AUC. A classifier comparison and feature-level latency analysis are also provided to clarify the accuracy–cost tradeoff and deployment feasibility. The results indicate that the proposed method can provide an interpretable and lightweight solution for electric bicycle controllers, battery management systems (BMSs), and onboard safety-monitoring applications. Full article

Background: Tandem leucine-rich repeats (LRRs) are typically classified into eleven types; however, several variant motifs have also been reported. Here, we identified new LRR variants that exhibit dual characteristics of two distinct types. We investigated how the dual characteristics influence the structure and function of LRRs. Methods: We conducted sequence similarity searches using the protein database and analyzed sequence features. We also characterized the structural features of these LRR variant motifs using solved structures and AlphaFold models and investigated their potential biological functions through domain analysis. Results: Of the identified 3222 proteins, approximately 60% originate from the bacterial PVC superphylum. The variants were classified into two groups: one defined by the consensus sequence LxxLxLxx(C/T)xzI TDxxLxx(L/F)xx(L/C)xx, and the other by LxxLxLxxCxxI TDxxLxxLxxLP (where “z” denotes a deletion). The LRRs highly similar to the variants are occasionally observed in solved structures and comprise three types of super-secondary structures (SSSs): β-strand–α-helix adjoining a 3(10)-helix–β-strand, β-strand–3(10)-helix–β-strand, and β-strand–3(10)-helix adjoining an α-helix–β-strand. The AlphaFold models adopt these SSSs and, in addition, include the SSS of the β–α–β motif. Functional annotation identified kinase and F-box domains in a subset of these LRR proteins. Conclusions: The coexistence of these four SSSs and the high frequency of the first SSS appear to reflect the dual characteristics of the LRR variants. The LRR variant-containing proteins suggest potential roles in bacterial immunity and ubiquitination. The present findings expand the structural diversity of LRR proteins and provide new insights into their functional roles. Full article

The internal quality of watermelons has garnered extensive attention. Conventional destructive quality detection for watermelons causes fruit loss, while existing acoustic techniques often rely on a single evaluation index. To address these issues, this study proposes a non-destructive method for comprehensive watermelon quality detection using acoustic and vibration signals. Signals from two watermelon varieties were collected under impact excitation to extract six time-domain and frequency-domain features. Factor Analysis of Mixed Data (FAMD) was employed to integrate ripeness, Soluble Solids Content (SSC), firmness, and sensory scores into a Comprehensive Quality Index (CQI), categorizing samples into High-Quality, Medium-Quality, and Low-Quality groups. Following physically constrained data augmentation to mitigate small sample size and class imbalance, an Extremely Randomized Trees (Extra-Trees) model was constructed. Results demonstrate that the Extra-Trees model achieved an overall testing accuracy of 0.92, with recall rates of 0.93 and 1.00 for Low-Quality and High-Quality watermelons, respectively. Recognition for Medium-Quality samples was lower due to overlapping physical and acoustic characteristics. Ultimately, this system aligns with actual consumer demands, providing technical support for low-cost, portable, and non-destructive watermelon grading. Full article

Lightweight polymer composites with effective electromagnetic interference (EMI) shielding are of increasing interest for advanced electronic and aerospace applications; however, conventional glass fiber-reinforced polymers (GFRPs) exhibit inherently low electrical conductivity, limiting their shielding performance. In this study, a hierarchical hybrid conductive architecture was developed by integrating graphene-coated multiaxial glass fiber fabrics with carbon black (CB)-reinforced epoxy matrices to enhance EMI shielding behavior in the X-band (8–12 GHz). Graphene coatings were deposited onto glass fibers via a surfactant-assisted ultrasonic dispersion method, while carbon black (0–1 wt.%) was incorporated into the epoxy matrix using ultrasonication-assisted mixing. Multilayer composites were fabricated using a vacuum bagging process. X-ray diffraction analysis revealed that the composites retained a predominantly amorphous epoxy/glass fiber matrix while exhibiting broad carbon-related diffraction features associated with disordered graphitic domains. Electrical conductivity measurements indicated that all composites remained in the insulating regime (~10⁻⁹ S/m), suggesting that a fully interconnected conductive network was not established within the investigated filler range. Despite the absence of a continuous conductive network, measurable EMI shielding performance was achieved. The composite containing 0.25 wt.% CB exhibited the highest shielding effectiveness, reaching approximately 12 dB at ~11.2 GHz. Analysis of the shielding contributions showed that absorption contributions (SEA) were consistently higher than reflection contributions (SER) across the studied frequency range. Morphological observations revealed that well-dispersed CB at low loading facilitated the formation of localized conductive domains that may contribute to tunneling-assisted polarization and interfacial charge accumulation. At higher CB contents, particle agglomeration reduced dispersion quality and limited effective pathway formation, while dynamic mechanical analysis indicated enhanced stiffness at low CB loading. FTIR results confirmed the absence of new chemical bonding, indicating that CB acts as a physically dispersed conductive filler. Overall, the results show that effective EMI shielding can be achieved in electrically insulating composites through the combined effect of hierarchical structural design and localized conductive features. This approach provides a practical pathway for developing lightweight EMI shielding materials with controlled filler loading and preserved structural integrity for aerospace and electronic applications. Full article

Natural polysaccharides from Siegesbeckia orientalis L. have been reported to exhibit promising bioactivities. To enhance extraction efficiency, low-temperature plasma-assisted extraction was optimized for S. orientalis L. polysaccharides using single-factor experiments and response surface methodology. Column chromatography purified a homogeneous SIE-III fraction, followed by structural characterization. Optimal parameters were 80 kV discharge voltage, 153 Hz frequency, and 109 s treatment time, under which the polysaccharide yield reached 15.68%, significantly higher than that of the conventional hot water extraction method. Plasma treatment loosened the raw material’s surface, potentially facilitating polysaccharide release. SIE-III had a molecular weight of 20.831 kDa and comprised mainly galactose (51.7%), rhamnose (19.1%), arabinose (11.3%), and galacturonic acid (9.9%). It featured typical rhamnogalacturonan-I (RG-I) domains and a triple-helix conformation. Fourier transform infrared spectroscopy and nuclear magnetic resonance confirmed both α- and β- glycosidic linkages, and methylation analysis revealed a highly branched →3,4)-Galp-(1→ structure. This study provides an effective extraction method for plant polysaccharides and valuable insights into their potential applications in the food and other industries. Full article

As autonomous driving technology expands from structured urban roads to unstructured outdoor environments, precise understanding of complex terrain has become a critical requirement for ensuring safe vehicle navigation. However, outdoor environments are characterized by high dynamics, drastic illumination variations, ambiguous category boundaries, and prohibitive annotation costs, making traditional supervised learning methods that rely on large amounts of pixel-level annotations difficult to generalize. In this paper, we propose a novel dual-path bidirectional interactive encoder, termed LapDINO, that effectively combines the strong semantic generalization capability of the self-supervised foundation model DINOv3 with the multi-scale frequency analysis capacity of the Laplacian pyramid. Specifically, we leverage DINOv3 to extract global semantic features as a “semantic map”, while simultaneously obtaining multi-scale high-frequency details through Laplacian pyramid decomposition as “structural contours”. Building upon this, we design a bidirectional cross-attention fusion mechanism that enables dynamic interaction and mutual refinement between semantic information and geometric details. Furthermore, we introduce a multi-branch attention enhancement module that extracts pyramid features from three complementary perspectives. To address domain shift, we design lightweight visual adapters that enable efficient fine-tuning of the frozen DINOv3 backbone. Finally, we construct two off-road terrain segmentation datasets, VOTD and VOCD, to facilitate research in this domain. Experimental results demonstrate that the proposed method achieves state-of-the-art performance, striking an optimal balance between accuracy and computational efficiency, thereby providing a robust and efficient engineering solution for terrain perception in off-road environments. Full article

Electroencephalographic (EEG) signals have been extensively studied for emotion detection and, more recently, as an alternative for biometric identification and authentication. Biometric methods based on physiological signals are a non-conventional approach for personal identification, and their study is currently considered an open research field. However, EEG-based biometric systems face several challenges, including the influence of emotional states, which can affect their performance. This study evaluates the effect of emotional states on the performance of an EEG-based biometric system. Four widely used databases for biometrics and emotion recognition (DEAP, MAHNOB, SEED, and LUMED-2) were selected for analysis. Feature extraction was performed using multiple strategies in the time, frequency, and time–frequency domains. The performance of various classifiers—support vector machine (SVM), random forest (RF), artificial neural networks (ANN), and k-nearest neighbors (K-NN)—was evaluated separately. Furthermore, stacking was used as a classifier fusion method. Explicit modeling of emotional states contributed to improving classifier performance. The best model based on classifier fusion achieved an accuracy of 95.73 ± 1.83%. These results indicate that incorporating information about emotional state into EEG-based biometric systems can contribute to the development of more robust and realistic identification solutions. Full article

In impact testing, evaluating multiple-impact signals is critical for verifying whether a test setup can reproduce penetration response signals and ensure reliable results. To overcome the limitations of traditional methods, including incomplete indicator coverage, subjective weighting, and poor consistency, this study proposes a fuzzy comprehensive evaluation (FCE) framework based on time–frequency features and combined weighting. Using multi-layer penetration response signals as the matching target, a multidimensional indicator system covering time-domain features, frequency-domain features, and signal quality and stability is established. A combined weighting method integrating AHP, EWM, and CRITIC is then developed, and subjective and objective weights are fused using the geometric mean method. A fuzzy comprehensive evaluation model is used to quantify the matching degrees of multiple sets of multiple-impact signals, and robustness is verified through weight consistency tests and sensitivity analysis. The results show that the evaluated signal sets are rated “Excellent”. Under reasonable weight combinations, the probability of obtaining an “Excellent” result reaches 99.94%, and the maximum variation caused by a ±10% perturbation in a single indicator weight is only 0.0087. The proposed framework provides a practical tool for evaluating multi-layer penetration response simulations and can be extended to other complex dynamic signal-matching problems. Full article

Underwater detection is crucial for the autonomous operation of Autonomous Underwater Vehicles (AUVs). However, underwater environments pose significant challenges, including severe image degradation, complex target deformation, and densely distributed small objects. Most existing methods treat image enhancement as an independent preprocessing module and rely on fixed-shape convolution kernels for feature extraction, which often leads to inconsistent optimization objectives and limited capability in handling irregular targets and fine-grained small-object details. To address these issues, we propose an End-to-End Adaptive Underwater Detection framework (E2E-AUD). Specifically, a lightweight image enhancement module, UnitModule, is embedded into the detection network so that enhancement can be jointly optimized with detection and directly serve downstream feature learning. In addition, linear deformable convolution (LDConv) is introduced into the backbone to adaptively model polymorphic targets, while Haar wavelet downsampling (HWD) is adopted to preserve boundary and texture information through frequency-domain analysis. Experiments on the DUO and URPC datasets demonstrate that E2E-AUD achieves superior performance over both general-purpose and underwater-specific detectors. Specifically, on the DUO dataset, our model reaches 86.2% mAP50 and 67.8% mAP50-95, outperforming the recent YOLOv12 by 3.0% and 2.7%, respectively. On the highly turbid URPC dataset, it achieves 84.3% mAP50 and 50.8% mAP50-95, surpassing the competitive underwater-specific detector LEFEN by notable margins in strict localization metrics. Furthermore, E2E-AUD maintains a real-time inference speed of 21.8 FPS with highly constrained computational complexity (9.4 GFLOPs), proving its exceptional balance between detection accuracy and deployment efficiency compared to previous methods. Full article

Time series anomaly detection aims to identify deviations from the normal distribution of temporal data. Reconstruction error is a natural and practical anomaly criterion, and using reconstruction error as an anomaly criterion is a well-established and practical paradigm. However, existing reconstruction-based methods often fail to capture complex structures in high-dimensional time series data and typically lack in-depth analysis of periodicity, limiting detection accuracy. To address these challenges, we propose ReDTF-AD, a novel approach that integrates reconstruction error with time–frequency fusion. Specifically, the input series is decomposed into seasonal and trend components. For seasonal components, we designed a time–frequency fusion block (TFFB) to enhance frequency-domain features while preserving residual time-domain information, ultimately achieving the fusion of time–frequency information. The Top-k transformation converts 1D sequences into 2D representations based on the periodicity of time series data, enabling deeper analysis of intra-period and inter-week variations through our newly proposed Split Concat Block (SCBlock). For the trend term, a linear module captures long-term patterns. In the unsupervised time series anomaly detection experiments based on reconstruction error, ReDTF-AD shows competitive performance on three public benchmark datasets. Full article

Background: Rare neurological diseases are challenging to diagnose from brain MRI because of their low prevalence, heterogeneous imaging patterns, and limited annotated datasets. Deep learning may support image-level recognition, but results from curated datasets without complete patient-level identifiers require cautious interpretation. Objectives: This study proposes RareNeuroXNet, a frequency-aware multi-branch attention framework for image-level classification of rare neurological diseases from brain MRI. The objective was to assess whether combining global anatomical, local fine-grained, and frequency-domain representations improves benchmark performance, calibration, and interpretability. Methods: RareNeuroXNet uses three complementary branches: a global branch for whole-image representation, a local branch for regional feature extraction, and an FFT magnitude-based frequency branch. Features are refined using CBAM attention, fused, and classified through a fully connected head. The model was evaluated on a balanced curated dataset with five rare neurological disease classes using five-fold cross-validation, ablation analysis, calibration metrics, internal baseline comparison, paired testing against DenseNet121 local-only, and Grad-CAM visualization. MCND was also used as a complementary cross-dataset neurological MRI benchmark, not as same-task external validation. Results: RareNeuroXNet achieved strong image-level internal benchmark performance, with accuracy of $0.9924\pm 0.0061$, macro F1-score of $0.9924\pm 0.0061$, macro AUROC of $0.9998\pm 0.0002$, and macro AUPR of $0.9992\pm 0.0007$. Calibration was favorable, with ECE of $0.0052\pm 0.0029$ and NLL of $0.0276\pm 0.0159$. Ablation results showed that the local branch was the dominant contributor, while FFT and CBAM provided supportive refinement. Compared with DenseNet121 local-only, RareNeuroXNet showed modest classification gains and clearer calibration improvements. Conclusions: RareNeuroXNet demonstrated strong controlled image-level benchmark performance with high discrimination, stable cross-validation behavior, favorable calibration, and Grad-CAM interpretability. However, possible correlated slices, duplicate images, or subject overlap cannot be excluded. Future work should use patient-level, same-task, multi-center external validation and 3D multimodal MRI analysis. Full article

Rotor imbalance and abnormal vibration are classical operating conditions in rotating machinery and can often be identified by conventional vibration analysis. However, the development of low-power, self-powered, and distributed sensing nodes remains important for long-term condition monitoring, particularly in scenarios where external power supply, wiring, and maintenance are constrained. Existing vibration sensors, including piezoelectric and capacitive types, are constrained by power consumption and degraded performance under low-frequency and weak excitation. To address this issue, a magnetically repulsive cushion triboelectric nanogenerator (MRCT) is proposed to enable self-powered vibration sensing. The magnetic-repulsion cushion allows the upper friction layer to undergo stable contact–separation motion under a non-contact restoring force, while the microstructured strip electrode array (MSEA) enhances the triboelectric output and signal stability. A hybrid convolutional neural network–gated recurrent unit (CNN-GRU) deep-learning model is employed to extract time-domain and frequency-domain features from the collected signals, enabling real-time identification of rotor vibration amplitude, frequency, and imbalance weight. Experimental results show that the MRCT provides stable output, a high signal-to-noise ratio, and an identification accuracy above 98% for predefined rotor imbalance-weight states under laboratory conditions. In addition, a shaft-misalignment-related abnormal vibration condition was examined on the motor platform. The corresponding time-domain and frequency-domain analyses show that the MRCT voltage signal exhibits distinguishable signal variations under normal and misalignment-related conditions, including spectral changes around the 2× rotational frequency. A laboratory-scale AIoT-oriented demonstration further verifies the feasibility of integrating MRCT signal acquisition, CNN-GRU inference, wireless transmission, and GUI-based visualization. It should be noted that the present work mainly focuses on imbalance-state recognition, while the misalignment-related experiment provides an additional sensor-response verification. Broader validation involving mechanical looseness, bearing defects, variable-speed operation, cross-machine testing, and long-term industrial conditions remains necessary. Full article

This paper proposes a two-stage hybrid framework for biosignal quality validation that produces beat-level or segment-level labels for real-time filtering and offline dataset curation. The framework is quantitatively validated exclusively on ECG data. Its modular architecture is designed to extend to further non-stationary periodic biomedical time-series signals including photoplethysmography (PPG), impedance cardiography (ICG), phonocardiography (PCG), electromyography (EMG), and electroencephalography (EEG) through modality-specific parameter adaptation; however, this broader applicability currently reflects architectural extensibility rather than experimentally validated performance. A prerequisite is synchronized acquisition of the primary biosignal together with inertial motion sensing (IMU/accelerometer) and electrode impedance or lead-off status, with the IMU positioned near the sensing electrodes. The first stage performs sensor-integrity gating to reject intervals corrupted by motion or poor electrode contact. The second stage applies software signal quality indices to the remaining beats, including physiological plausibility constraints (R to R peaks analysis), DTW-based morphological consistency against adaptive templates, frequency domain SNR estimation, and baseline wander quantification. This study systematically evaluates and compares the classification performance of six complementary sensor-level and software-based signal quality assessment methods. When integrated within the proposed hybrid framework, validation against expert-annotated ECG quality labels from 20 healthy participants demonstrates high methodological classification accuracy (98.1%), achieving approximately a 98% F1-score, 99% sensitivity, and 97% specificity. Prospective validation on patient populations with cardiovascular pathology is identified as a necessary step toward clinical deployment. This modular approach improves the reliability of downstream analysis by preventing corrupted data from entering feature extraction and model training pipelines, enabling more stable physiological monitoring in free-living conditions, reducing false alarms in continuous monitoring applications, and generating higher-quality datasets for AI-based diagnostic systems. Full article

Background: Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) interpolation and segmentation are critical for clinical diagnosis, anatomical quantification and personalized treatment. Most existing methods perform these two tasks separately, leading to computational redundancy and insufficient mining of shared spatial features. This study aims to construct an integrated multi-task learning framework for the synchronous processing of medical image interpolation and segmentation. Methods: We propose a unified multi-task framework named TASC-SwinMT for joint interpolation and multi-frame segmentation of CT and MRI images. It employs a shared SwinUNet encoder to extract general spatial features, matched with two task-specific decoders for frame prediction and mask generation. Three functional modules are designed for cross-task synergistic learning, and a dynamic multi-task loss function is used to balance objective optimization. Experiments are performed on Medical Segmentation Decathlon Task02_Heart and Task06_Lung datasets. Results: Our method outperforms baseline models and ablation variants in both tasks with outstanding accuracy and significantly reduced computational overhead. It exhibits superior performance in lesion boundary depiction, small object segmentation and inter-slice consistency, and anatomical prior constraints with frequency-domain modeling further enhance prediction quality. Conclusions: The cross-task feature sharing and joint optimization strategy are validated effective. The proposed TASC-SwinMT framework has favorable stability and generalization ability, providing a reliable solution for clinical medical image analysis. Full article

As unmanned aerial vehicles (UAVs) are increasingly deployed in various fields, their flight safety has become a critical issue. However, limited onboard sensing and computing resources make it difficult to perform intelligent fault monitoring and diagnosis directly on UAVs. To explore an offboard alternative, this paper investigates a drone nest vibration analysis based fault diagnosis framework for a multirotor UAV rotor system using vibration signals measured from a laboratory-scale simulated drone nest. A simplified coupled dynamic model of the UAV–drone nest system is established to analyze the transmission mechanism of rotor fault-induced vibration and to explain the observability of fault-related frequency components under the tested configuration. Considering the weak and attenuated characteristics of the nest-side vibration signals, a multi-domain feature fusion and multi-task learning network is developed to integrate time-domain, frequency-domain, and envelope-spectrum information while jointly learning fault type and rotational speed. Comparative experiments on the constructed quadrotor–drone nest test platform are conducted to validate the feasibility and effectiveness of the proposed method under the tested operating conditions. Full article