Sensors最新文献

The fusion of low-resolution hyperspectral images (LR-HSI) with high-resolution multispectral images (HR-MSI) provides a cost-effective approach to obtaining high-resolution hyperspectral images (HR-HSI). Existing methods primarily based on convolutional neural networks (CNNs) struggle to capture global features and do not adequately address the significant scale and spectral resolution differences between LR-HSI and HR-MSI. To tackle these challenges, our novel FCSwinU network leverages the spectral fast Fourier convolution (SFFC) module for spectral feature extraction and utilizes the Swin Transformer's self-attention mechanism for multi-scale global feature fusion. FCSwinU employs a UNet-like encoder-decoder framework to effectively merge spatiospectral features. The encoder integrates the Swin Transformer feature abstraction module (SwinTFAM) to encode pixel correlations and perform multi-scale transformations, facilitating the adaptive fusion of hyperspectral and multispectral data. The decoder then employs the Swin Transformer feature reconstruction module (SwinTFRM) to reconstruct the fused features, restoring the original image dimensions and ensuring the precise recovery of spatial and spectral details. Experimental results from three benchmark datasets and a real-world dataset robustly validate the superior performance of our method in both visual representation and quantitative assessment compared to existing fusion methods.

In the era of digital music proliferation, music genre classification has become a crucial task in music information retrieval. This paper proposes a novel channel-aware convolutional neural network (ECAS-CNN) designed to enhance the efficiency and accuracy of music genre recognition. By integrating an adaptive channel attention mechanism (ECA module) within the convolutional layers, the network significantly improves the extraction of key musical features. Extensive experiments were conducted on the GTZAN dataset, comparing the proposed ECAS-CNN with traditional convolutional neural networks. The results demonstrate that ECAS-CNN outperforms conventional methods across various performance metrics, including accuracy, precision, recall, and F1-score, particularly in handling complex musical features. This study validates the potential of ECAS-CNN in the domain of music genre classification and offers new insights for future research and applications.

Due to the complex intersection and control of multiple structural systems, the hydrogeological conditions of the Laiyuan Basin in China are complex. The depth of research on the relationship between geological structure and groundwater migration needs to be improved. The supply relationship of each aquifer is still uncertain. This paper systematically conducts research on the characteristics of hydrogen and oxygen isotopes, and combines magnetotelluric impedance tensor decomposition and two-dimensional fine inversion technology to carry out fine exploration of the strata and structures in the Laiyuan Basin, as well as comprehensive characteristics of groundwater migration and replenishment. The results indicate the following: (i) The hydrogen and oxygen values all fall near the local meteoric water line, indicating that precipitation is the main groundwater recharge source. (ii) The excess deuterium decreased gradually from karst mountain to basin, and karst water and pore water experienced different flow processes. (iii) The structure characteristics of three main runoff channels are described by MT fine processing and inversion techniques. Finally, it is concluded that limestone water moved from the recharge to the discharge area, mixed with the deep dolomite water along the fault under the control of fault F2, and eventually rose to the surface of the unconsolidated sediment blocked by fault F1 to emerge into an ascending spring.

The objective of this study was to assess the accuracy of virtual reality (VR) technology in replicating real-life environments for the adoption of appropriate human postures and forces. Despite the widespread implementation of VR in various applications, there is a lack of research evaluating the accuracy of human postures and sensory aspects in the VR environment compared to real-life scenarios. A total of twenty-two student participants were recruited for this study, which involved a common lifting task. Two specific poses were identified as having potentially excessive forces exerted on the lower back. By comparing the angles of seven anatomical joints in both the real environment and the VR environment at each pose, we observed that depth perception may influence posture adoption in the VR setting. Moreover, the presence of a physical load applied to both hands significantly influenced the postures adopted by participants compared to those in the VR environment. These deviations in postures directly led to significant differences in predicted spinal forces exerted on the lower back, which in turn could result in inaccurate assessments of injury risks and the design of injury prevention programs. Therefore, it is crucial to understand the accuracy of VR technology as a substitute for real-life environments.

Underwater simultaneous localization and mapping (SLAM) is essential for effectively navigating and mapping underwater environments; however, traditional SLAM systems have limitations due to restricted vision and the constantly changing conditions of the underwater environment. This study thoroughly examined the underwater SLAM technology, particularly emphasizing the incorporation of deep learning methods to improve performance. We analyzed the advancements made in underwater SLAM algorithms. We explored the principles behind SLAM and deep learning techniques, examining how these methods tackle the specific difficulties encountered in underwater environments. The main contributions of this work are a thorough assessment of the research into the use of deep learning in underwater image processing and perception and a comparison study of standard and deep learning-based SLAM systems. This paper emphasizes specific deep learning techniques, including generative adversarial networks (GANs), convolutional neural networks (CNNs), long short-term memory (LSTM) networks, and other advanced methods to enhance feature extraction, data fusion, scene understanding, etc. This study highlights the potential of deep learning in overcoming the constraints of traditional underwater SLAM methods, providing fresh opportunities for exploration and industrial use.

In this paper, we address the issues of insufficient accuracy and frequent identity switching in the multi-target tracking algorithm DeepSORT by proposing two improvement strategies. First, we optimize the appearance feature extraction process by training a lightweight appearance extraction network (OSNet) on a vehicle re-identification dataset. This makes the appearance features better suited for the vehicle tracking model required in our paper. Second, we improve the metric of motion features by using the original IOU distance metric or GIOU metrics. The optimized tracking algorithm using GIOU achieves effective improvements in tracking precision and accuracy. The experimental results show that the improved vehicle tracking models MOTA and IDF1 are enhanced by 4.6% and 5.9%, respectively. This allows for the stable tracking of vehicles and reduces the occurrence of identity switching phenomenon to a certain extent.

This study proposes a wearable gait assessment method using inertial measurement units (IMUs) to evaluate gait ability in daily environments. By focusing on the estimation of the margin of stability (MoS), a key kinematic stability parameter, a method using a convolutional neural network, was developed to estimate the MoS from IMU acceleration time-series data. The relationship between MoS and other stability indices, such as the Lyapunov exponent and the multi-site time-series (MSTS) index, using data from five IMU sensors placed on various body parts was also examined. To simulate diverse gait conditions, treadmill speed was varied, and a knee-ankle-foot orthosis was used to restrict left knee extension, inducing gait asymmetry. The model achieved over 90% accuracy in classifying MoS in both forward and lateral directions using three-axis acceleration data from the IMUs. However, the correlation between MoS and the Lyapunov exponent or MSTS index was weak, suggesting that these indices may capture different aspects of gait stability.

This paper investigates methods that leverage physical contact between a robot's structure and its environment to enhance task performance, with a primary emphasis on improving precision. Two main approaches are examined: solving the inverse kinematics problem and employing quadratic programming, which offers computational efficiency by utilizing forward kinematics. Additionally, geometrical methods are explored to simplify robot assembly and reduce the complexity of control calculations. These approaches are implemented on a physical robotic platform and evaluated in real-time applications to assess their effectiveness. Through experimental evaluation, this study aims to understand how environmental contact can be utilized to enhance performance across various conditions, offering valuable insights for practical applications in robotics.

Sleep, notably active sleep (AS) and quiet sleep (QS), plays a pivotal role in the brain development and gradual maturation of (pre) term infants. Monitoring their sleep patterns is imperative, as it can serve as a tool in promoting neurological maturation and well-being, particularly important in preterm infants who are at an increased risk of immature brain development. An accurate classification of neonatal sleep states can contribute to optimizing treatments for high-risk infants, with respiratory rate (RR) and heart rate (HR) serving as key components in sleep assessment systems for neonates. Recent studies have demonstrated the feasibility of extracting both RR and HR using near-infrared spectroscopy (NIRS) in neonates. This study introduces a comprehensive sleep classification approach leveraging high-frequency NIRS signals recorded at a sampling rate of 100 Hz from a cohort of nine preterm infants admitted to a neonatal intensive care unit. Eight distinct features were extracted from the raw NIRS signals, including HR, RR, motion-related parameters, and proxies for neural activity. These features served as inputs for a deep convolutional neural network (CNN) model designed for the classification of AS and QS sleep states. The performance of the proposed CNN model was evaluated using two cross-validation approaches: ten-fold cross-validation of data pooling and five-fold cross-validation, where each fold contains two independently recorded NIRS data. The accuracy, balanced accuracy, F1-score, Kappa, and AUC-ROC (Area Under the Curve of the Receiver Operating Characteristic) were employed to assess the classifier performance. In addition, comparative analyses against six benchmark classifiers, comprising K-Nearest Neighbors, Naive Bayes, Support Vector Machines, Random Forest (RF), AdaBoost, and XGBoost (XGB), were conducted. Our results reveal the CNN model's superior performance, achieving an average accuracy of 88%, a balanced accuracy of 94%, an F1-score of 91%, Kappa of 95%, and an AUC-ROC of 96% in data pooling cross-validation. Furthermore, in both cross-validation methods, RF and XGB demonstrated accuracy levels closely comparable to the CNN classifier. These findings underscore the feasibility of leveraging high-frequency NIRS data, coupled with NIRS-based HR and RR extraction, for assessing sleep states in neonates, even in an intensive care setting. The user-friendliness, portability, and reduced sensor complexity of the approach suggest its potential applications in various less-demanding settings. This research thus presents a promising avenue for advancing neonatal sleep assessment and its implications for infant health and development.

The stratum corneum (SC) forms the outermost layer of the skin, playing a critical role in preventing water loss and protecting against external biological and chemical threats. Approximately 90% of the SC consists of large, flat corneocytes, yet its barrier function primarily relies on the intercellular lipid matrix that surrounds these cells. Traditional methods for characterizing these lipids, such as Fourier transform infrared spectroscopy (FTIR), typically involve macroscopic analysis using attenuated total reflection (ATR) techniques. In this study, we introduce a novel approach for investigating SC samples at a microscopic level to gain detailed chemical insights and assess sample heterogeneity. Special emphasis is placed on advanced hyperspectral data pre-processing to ensure the accuracy and reliability of the results. We also evaluate methods for filtering out spectral data that significantly deviate from the mean and analyze the extracted mean spectra, the intensities of specific infrared peaks, and their ratios. The novelty of this work lies in its microscopic approach to analyzing the SC lipid matrix, diverging from the traditional macroscopic FTIR-ATR methods. By focusing on hyperspectral imaging and developing robust pre-processing techniques, this study provides more localized, high-resolution chemical insights. This microscopic perspective opens up the possibility of detecting subtle heterogeneities within the skin's lipid matrix, offering deeper, previously unattainable understanding of the SC's barrier function. Additionally, the exploration of spectral filtering methods enhances the precision of the analysis, paving the way for more refined and reliable investigations of skin structure and behavior in future research.