Sensors最新文献

This study explores the potential of accelerometer-based gait analysis as a non-invasive approach for predicting cognitive impairment in older adults. A total of 75 participants (61.3% female; mean age: 78.9 years), including cognitively normal individuals and patients with dementia, were enrolled. Walking data were collected using a six-axis waist-worn accelerometer during self-paced locomotion. Allan variance (AVAR), a robust statistical measure of frequency stability, was applied to characterize gait dynamics. AVAR-derived features, combined with participant age, were used as inputs to machine learning models, logistic regression and Light Gradient Boosting Machine (LightGBM) for classifying cognitive status based on Mini-Mental State Examination (MMSE) scores. LightGBM achieved superior performance (AUC = 0.92) compared to logistic regression (AUC = 0.85). Although mild cognitive impairment (MCI) cases were grouped with cognitively normal participants, gait-based classification revealed that MCI individuals exhibited patterns more similar to those with cognitive impairment. These results suggest that AVAR-based gait features are promising for early detection of cognitive decline in older adults.

Campus stairwells, characterized by their crowded nature during certain short periods of time, present a high risk for falls that can lead to dangerous stampedes. Accurate fall detection is crucial for preventing such accidents. However, existing research lacks a detection model that balances high precision with lightweight design and lacks on-site experimental validation to assess practical feasibility. This study addresses these gaps by proposing an enhanced fall recognition model based on YOLOv7, validated through on-site experiments. A dataset on campus stairwell falls was established, capturing diverse stairwell personnel behaviors. Four YOLOv7 improvement schemes were proposed, and numerical comparison experiments identified the best-performing model, combining DO-DConv and Slim-Neck modules. This model achieved an average precision (mAP) of 88.1%, 2.41% higher than the traditional YOLOv7, while reducing GFLOPs from 105.2 to 38.2 and cutting training time by 4 h. A field experiment conducted with 22 groups of participants under small-scale populations and varying lighting conditions preliminarily confirmed that the model's accuracy is within an acceptable range. The experimental results also analyzed the changes in detection confidence across different population sizes and lighting conditions, offering valuable insights for further model improvement and its practical applications.

This paper describes an ultra-high-speed monolithic global shutter CMOS image sensor capable of continuous motion capture at 326,000 fps with a resolution of 640 × 480 pixels. The performance is enabled by a novel combination of pixel technology and circuit techniques. The highly sensitive BSI pixel with a 52 μm pitch employs a fully depleted substrate to facilitate rapid photocarrier transport. In-pixel voltage mode storage enables pipelined readout, while in-pixel analog CDS provides low noise with minimal impact on readout speed. The sensor achieves an equivalent row time of 6.4 ns through separate top and bottom readout together with multiple parallel ADCs per column. Independent row drivers on both the left and right sides ensure the global shutter accuracy needed for the minimum exposure time of 59 ns. The dynamic range is enhanced by on-chip reduction in FPN and by PTC-based data compression. The sensor delivers a throughput of 100 Gpix/sec, transferred off chip via 128 CML channels operating at 6.6 Gbps each. The device is fabricated using a 130 nm monolithic CIS process with BSI postprocessing and is in series production.

In recent years, video-based monitoring systems have been widely adopted across multiple domains and have become particularly vital in geohazard monitoring and early warning. These systems overcome the inherent limitations of conventional monitoring techniques by enabling real-time, non-contact, and intuitive visual observation of geologically hazardous sites. With the integration of machine learning and other advanced analytical methods, video-based systems can process and interpret image data in real time, thereby supporting rapid detection and timely early warning of potential geohazards. This substantially improves both the efficiency and accuracy of monitoring efforts. Drawing on domestic and international research, this article provides a comprehensive review of video-based monitoring technologies, machine learning-driven video image processing, and multi-source data fusion approaches. It systematically summarizes their underlying technical principles and applications in geohazard monitoring and early warning, and offers an in-depth analysis of their practical advantages and future development trends. This review aims to serve as a valuable reference for advancing research and innovation in this field.

This article proposes a novel passive localization framework that leverages detection results from existing distributed radio detectors. The intuition behind this solution is to combine positive (signal detected) and negative (signal not detected) detection results with environmental data to refine localization estimates. Its novelty lies in providing a comprehensive, multi-dimensional framework for cooperative localization that enhances situational awareness by leveraging existing spectrum-monitoring capabilities. The proposed approach provides an additional functionality for a network of nodes monitoring spectral resources. It allows the transmitter's location to be estimated based on the detection results of individual nodes. The unquestionable advantage of the proposed solution is that it does not require extra equipment or increased monitoring time. The developed method supports broad operational activities, e.g., tracking of authorized and unauthorized entities, and jammer localization. Using the proposed approach, one can increase efficiency in a given operational environment, and jammer localization. Using the proposed approach, one can increase efficiency in a given operational environment and situational awareness in a cognitive radio network. Furthermore, the experimental results of the estimation algorithm for an exemplary urban area indicate the legitimacy of a cooperative approach to the problem.

5G technology represents a transformative shift in mobile communications, delivering improved ultra-low latency, data throughput, and the capacity to support huge device connectivity, surpassing the capabilities of LTE systems. As global telecommunication operators shift toward widespread 5G implementation, ensuring optimal network performance and intelligent resource management has become increasingly obvious. To address these challenges, this study explored the role of advanced clustering methods in optimizing cellular networks under heterogeneous and dynamic conditions. A systematic literature review (SLR) was conducted by analyzing 40 peer-reviewed and non-peer-reviewed studies selected from an initial collection of 500 papers retrieved from the Semantic Scholar Open Research Corpus. This review examines a diversity of clustering approaches, including spectral clustering with Bayesian non-parametric models and K-means, density-based clustering such as DBSCAN, and deep representation-based methods like Differential Evolution Memetic Clustering (DEMC) and Domain Adaptive Neighborhood Clustering via Entropy Optimization (DANCE). Key performance outcomes reported across studies include anomaly detection accuracy of up to 98.8%, delivery rate improvements of up to 89.4%, and handover prediction accuracy improvements of approximately 43%, particularly when clustering techniques are combined with machine learning models. In addition to summarizing their effectiveness, this review highlights methodological trends in clustering parameters, mechanisms, experimental setups, and quality metrics. The findings suggest that advanced clustering models play a crucial role in intelligent spectrum sensing, adaptive mobility management, and efficient resource allocation, thereby contributing meaningfully to the development of intelligent 5G/6G mobile network infrastructures.

To address active voltage control in photovoltaic (PV)-integrated distribution networks characterized by weak voltage support conditions, this paper proposes a multi-agent deep reinforcement learning (MADRL)-based coordinated control method for PV clusters. First, the voltage control problem is formulated as a decentralized partially observable Markov decision process (Dec-POMDP), and a centralized training with decentralized execution (CTDE) framework is adopted, enabling each inverter to make independent decisions based solely on local measurements during the execution phase. To balance voltage compliance with energy efficiency, two barrier functions are designed to reshape the reward function, introducing an adaptive penalization mechanism: a steeper gradient in violation region to accelerate voltage recovery to the nominal range, and a gentler gradient in the safe region to minimize excessive reactive regulation and power losses. Furthermore, six representative MADRL algorithms-COMA, IDDPG, MADDPG, MAPPO, SQDDPG, and MATD3-are employed to solve the active voltage control problem of the distribution network. Case studies based on a modified IEEE 33-bus system demonstrate that the proposed framework ensures voltage compliance while effectively reducing network losses. The MADDPG algorithm achieves a Controllability Ratio (CR) of 91.9% while maintaining power loss at approximately 0.0695 p.u., demonstrating superior convergence and robustness. Comparisons with optimal power flow (OPF) and droop control methods confirm that the proposed approach significantly improves voltage stability and energy efficiency under model-free and communication-constrained weak grid conditions.

This study introduces a sensor-centric cybersecurity framework for railway infrastructure that extends Failure Mode and Effects Analysis (FMEA) from traditional reliability evaluation into the domain of cyber-induced failures affecting data integrity, availability and authenticity. The contribution lies in bridging regulatory obligations of the NIS2 Directive with field-layer monitoring by enabling risk indicators to evolve dynamically rather than remain static documentation artefacts. The approach is demonstrated using a scenario-based dataset collected from approximately 250 trackside, rolling-stock, environmental and power-monitoring sensors deployed over a 25 km operational segment, with representative anomalies generated through controlled spoofing, replay and injection conditions. Risk was evaluated using RPN scores derived from Severity-Occurrence-Detectability scales, while anomaly-detection performance was observed through detection-latency variation, changes in RPN distribution, and qualitative responsiveness of timestamp-based alerts. Instead of presenting a fixed benchmark, the results show how evidence from real sensor streams can recalibrate O and D factors in near-real-time and reduce undetected exposure windows, enabling measurable compliance documentation aligned with NIS2 Article 21. The findings confirm that coupling FMEA with streaming telemetry creates a verifiable risk-evaluation loop and supports a transition toward continuous, evidence-driven cybersecurity governance in railway systems.

In this paper, we propose a novel speech emotion recognition model, named MAGTF-Net (Multi-scale Attention Graph Transformer Fusion Network), which addresses the challenges faced by traditional hand-crafted feature-based approaches in modeling complex emotional nuances and dynamic contextual dependencies. Although existing state-of-the-art methods have achieved improvements in recognition performance, they often fail to simultaneously capture both local acoustic features and global temporal structures, and they lack adaptability to variable-length speech utterances, thereby limiting their accuracy and robustness in recognizing complex emotional expressions. To tackle these challenges, we design a log-Mel spectrogram feature extraction branch that combines a Multi-scale Attention Graph (MAG) structure with a Transformer encoder, where the Transformer module adaptively performs dynamic modeling of speech sequences with varying lengths. In addition, a low-level descriptor (LLD) feature branch is introduced, where a multilayer perceptron (MLP) is employed for complementary feature modeling. The two feature branches are fused and subsequently classified through a fully connected layer, further enhancing the expressive capability of emotional representations. Moreover, a label-smoothing-enhanced cross-entropy loss function is adopted to improve the model's recognition performance on difficult-to-classify emotional categories. Experiments conducted on the IEMOCAP dataset demonstrate that MAGTF-Net achieves weighted accuracy (WA) and unweighted accuracy (UA) scores of 69.15% and 70.86%, respectively, outperforming several baseline models. Further ablation studies validate the significant contributions of each module in the Mel-spectrogram branch and the LLD feature branch to the overall performance improvement. The proposed method effectively integrates local, global, and multi-source feature information, significantly enhancing the recognition of complex emotional expressions and providing new theoretical and practical insights for the field of speech emotion recognition.

Predicting the fatigue lifespan of Twisted String Actuators (TSAs) is essential for improving the reliability of robotic and mechanical systems that rely on flexible transmission mechanisms. Traditional empirical approaches based on regression or Weibull distribution analysis have provided useful approximations, yet they often struggle to capture nonlinear dependencies and stochastic influences inherent to real-world fatigue behavior. This study introduces and compares four machine learning (ML) models-Linear Regression, Random Forest, XGBoost, and Gaussian Process Regression (GPR)-for predicting TSA lifespan under varying weight (W), number of strings (N), and diameter (D) conditions. Building upon this comparison, a hybrid physics-guided model is proposed by integrating an empirical fatigue life equation with an XGBoost residual-correction model. Experimental data collected from repetitive actuation tests (144 valid samples) served as the basis for training and validation. The hybrid model achieved an R² = 0.9856, RMSE = 5299.47 cycles, and MAE = 3329.67 cycles, outperforming standalone ML models in cross-validation consistency (CV R² = 0.9752). The results demonstrate that physics-informed learning yields superior interpretability and generalization even in limited-data regimes. These findings highlight the potential of hybrid empirical-ML modeling for component life prediction in robotic actuation systems, where experimental fatigue data are scarce and operating conditions vary.