Search

Abstract Specific emitter identification technology plays a very important role in spectrum resource management, wireless network security, cognitive radio etc. However, in complex electromagnetic environments, the variability and uncertainty of signals make it very difficult to extract representative feature representations of the signals. At the same time, the feature extraction capability of the recognition model is also a factor that needs to be considered. To address these issues, a wavelet residual neural network model based on attention mechanism is proposed for specific emitter identification. First, multi‐level wavelet decomposition is performed on all received signals to obtain their wavelet detail coefficients at different scales. Next, all the wavelet detail coefficients are used as the feature input for the attention‐based residual network, and perform parallel feature extraction at multi scales. Finally, the feature representation capability of all coefficients are compared, and the model's recognition results based on it are obtained. The recognition rates on the three datasets are 94.7%, 93.21%, and 86.1%, respectively, all of which are superior to recent state‐of‐the‐art algorithms. In addition, through ablation experiment, the validity of each component of the model has been verified.

Abstract Transformer tracking always takes paired template and search images as encoder input and conduct feature extraction and target‐search feature correlation by self and/or cross attention operations, thus the model complexity will grow quadratically with the number of input images. To alleviate the burden of this tracking paradigm and facilitate practical deployment of Transformer‐based trackers, we propose a dual pooling transformer tracking framework, dubbed as DPT, which consists of three components: a simple yet efficient spatiotemporal attention model (SAM), a mutual correlation pooling Transformer (MCPT) and a multiscale aggregation pooling Transformer (MAPT). SAM is designed to gracefully aggregates temporal dynamics and spatial appearance information of multi‐frame templates along space‐time dimensions. MCPT aims to capture multi‐scale pooled and correlated contextual features, which is followed by MAPT that aggregates multi‐scale features into a unified feature representation for tracking prediction. DPT tracker achieves AUC score of 69.5 on LaSOT and precision score of 82.8 on TrackingNet while maintaining a shorter sequence length of attention tokens, fewer parameters and FLOPs compared to existing state‐of‐the‐art (SOTA) Transformer tracking methods. Extensive experiments demonstrate that DPT tracker yields a strong real‐time tracking baseline with a good trade‐off between tracking performance and inference efficiency.

Abstract Lip‐reading deciphers speech by observing lip movements without relying on audio data. The rapid advancements in deep learning have significantly improved lip‐reading for both English and Chinese; however, research on dialects such as Cantonese remains scarce. Consequently, most Chinese lip‐reading datasets focus on Mandarin, with only a few addressing Cantonese. To bridge this gap, a sentence‐level Cantonese lip‐reading dataset, designated as Cantonese lip‐reading sentences are introduced, comprising over 500 unique speakers and more than 30,000 samples. To ensure alignment with real‐world scenarios, no restrictions are imposed on factors such as gender, age, posture, lighting conditions, or speech rate. A comprehensive description of the pipeline employed is provided for collecting and constructing the dataset and introduce an innovative visual frontend, 3D‐visual attention net. This frontend combines the advantages of convolution and self‐attention mechanisms to extract fine‐grained lip region features. These features are subsequently input into the conformer backend for temporal sequence modelling, achieving comparable performance on Chinese Mandarin lip reading dataset, lip reading sentences 2, lip reading sentences 3, and Cantonese lip‐reading sentences datasets. Benchmark tests on Cantonese lip‐reading sentences demonstrate the challenges it poses, providing a novel research foundation for dialect lip‐reading and fostering the advancement of Cantonese lip‐reading tasks.

Abstract Existing segmentation‐based methods have made considerable progress in arbitrarily shaped text detection due to the advantage of dealing with shape variation. However, there still exist challenges to detecting accurate text instances with dense layouts, inaccurate annotations, and complex backgrounds. Many recent works have focused on improving arbitrary boundary prediction, but it may be difficult to accurately distinguish each instance of dense layouts because their boundary pixels may be mistakenly classified to produce inaccurate results (i.e., adhesive texts) with inaccurate annotation and complex backgrounds. Considering the local and long‐range dependencies, this paper proposes an efficient text detector, namely T‐Skeleton, to obtain more reliable segmentation detections. In the spirit of object skeletonization, we introduce the text instance skeleton highlighting the semantically significant structure (similar to the skeleton of a fish) to explicitly capture the long‐range dependencies of text instances. The key idea of T‐Skeleton is to calibrate the coarse text proposals by embedding text instance skeletons to separate crowd texts accurately and robustly. We further design a channel attention module to enlarge the performance margin between T‐Skeleton and the segmentation baseline. Experimental results on four publicly available datasets show the superiority of T‐Skeleton in handling long and curved texts.

Abstract Amid growing interest in the precise detection of volatile organic compounds (VOCs) in industrial field, the demand for highly effective gas sensors is at an all‐time high. However, traditional sensors with their classic single‐output signal, bulky and complex integrated structure when forming array often involve complicated technology and high cost, limiting their widespread adoption. Here, this study introduces a novel approach, employing an integrated YSZ‐based (YSZ: yttria‐stabilized zirconia) mixed potential sensor equipped with a triple‐sensing electrode array, to efficiently detect and differentiate six types of VOCs gases. This innovative sensor integrates NiSb2O6, CuSb2O6, and MgSb2O6 sensing electrodes (SEs), which are sensitive to pentane, isoprene, n‐propanol, acetone, acetic acid, and formaldehyde gases. Through feature engineering based on intuitive spike‐based response values, it accentuates the distinct characteristics of every gas. Eventually, an average classification accuracy of 98.8% and an overall R‐squared error (R2) of 99.3% for concentration regression toward six target gases can be achieved, showcasing the potential to quantitatively distinguish between industrial hazardous VOCs gases.

Abstract Background Neurological disorders pose a significant health challenge, and their early detection is critical for effective treatment planning and prognosis. Traditional classification of neural disorders based on causes, symptoms, developmental stage, severity, and nervous system effects has limitations. Leveraging artificial intelligence (AI) and machine learning (ML) for pattern recognition provides a potent solution to address these challenges. Therefore, this study focuses on proposing an innovative approach—the Aggregated Pattern Classification Method (APCM)—for precise identification of neural disorder stages. Method The APCM was introduced to address prevalent issues in neural disorder detection, such as overfitting, robustness, and interoperability. This method utilizes aggregative patterns and classification learning functions to mitigate these challenges and enhance overall recognition accuracy, even in imbalanced data. The analysis involves neural images using observations from healthy individuals as a reference. Action response patterns from diverse inputs are mapped to identify similar features, establishing the disorder ratio. The stages are correlated based on available responses and associated neural data, with a preference for classification learning. This classification necessitates image and labeled data to prevent additional flaws in pattern recognition. Recognition and classification occur through multiple iterations, incorporating similar and diverse neural features. The learning process is finely tuned for minute classifications using labeled and unlabeled input data. Results The proposed APCM demonstrates notable achievements, with high pattern recognition (15.03%) and controlled classification errors (CEs) (10.61% less). The method effectively addresses overfitting, robustness, and interoperability issues, showcasing its potential as a powerful tool for detecting neural disorders at different stages. The ability to handle imbalanced data contributes to the overall success of the algorithm. Conclusion The APCM emerges as a promising and effective approach for identifying precise neural disorder stages. By leveraging AI and ML, the method successfully resolves key challenges in pattern recognition. The high pattern recognition and reduced CEs underscore the method's potential for clinical applications. However, it is essential to acknowledge the reliance on high‐quality neural image data, which may limit the generalizability of the approach. The proposed method allows future research to refine further and enhance its interpretability, providing valuable insights into neural disorder progression and underlying biological mechanisms.

Abstract Reconstruction‐based methods are commonly used in industrial visual anomaly detection. They rely on a well‐reconstructed normal mode of the model. However, it is difficult to manage the boundary of generalization. The strength of the model's generalization capability can directly affect the fidelity of the reconstruction, resulting in the occurrence of false positives. To address the above challenges, a novel dual branch reconstruction anomaly detection approach is proposed to control the model generalization capability at two dimensions. It reconstructs abnormal images into normal ones by resolution recovery and denoising branches. Detection results are generated from their comparison. In addition, an innovative channel adjustment module is introduced to improve information exchange between branches. It uses multiple dilated convolutions for interactions over different scales. Simulation experiments demonstrate that the method outperforms most inspection methods on the MVTec AD and MVTec 3D‐AD datasets. It also shows good results on the self‐generated automotive paint scratches dataset of this study.

Abstract In recent years, intelligent robots are extensively applied in the field of the industry and intelligent rehabilitation, wherein the human‐robot interaction (HRI) control strategy is a momentous part that needs to be ameliorated. Specially, the efficacy and robustness of the HRI control algorithm in the presence of unknown external disturbances deserve to be addressed. To deal with these urgent issues, in this study, artificial systems, computational experiments and a parallel execution intelligent control framework are constructed for the HRI control. The upper limb‐robotic exoskeleton system is re‐modelled as an artificial system. Depending on surface electromyogram‐based subject's active motion intention in the practical system, a non‐convex function activated anti‐disturbance zeroing neurodynamic (NC‐ADZND) controller is devised in the artificial system for parallel interaction and HRI control with the practical system. Furthermore, the linear activation function‐based zeroing neurodynamic (LAF‐ZND) controller and proportional‐derivative (posterior deltoid (PD)) controller are presented and compared. Theoretical results substantiate the global convergence and robustness of the proposed controller in the presence of different external disturbances. In addition, the simulation results verify that the NC‐ADZND controller is better than the LAF‐ZND and the PD controllers in respect of convergence order and anti‐disturbance characteristics.

Abstract In real‐world scenarios, pedestrian images often suffer from occlusion, where certain body features become invisible, making it challenging for existing methods to accurately identify pedestrians with the same ID. Traditional approaches typically focus on matching only the visible body parts, which can lead to misalignment when the occlusion patterns vary. To address this issue and alleviate misalignment in occluded pedestrian images, the authors propose a novel framework called body topology information generation and matching. The framework consists of two main modules: the body topology information generation module and the body topology information matching module. The body topology information generation module employs an adaptive detection mechanism and capsule generative adversarial network to restore a holistic pedestrian image while preserving the body topology information. The body topology information matching module leverages the restored holistic image from body topology information generation to overcome spatial misalignment and utilises cosine distance as the similarity measure for matching. By combining the body topology information generation and body topology information matching modules, the authors achieve consistency in the body topology information features of pedestrian images, ranging from restoration to retrieval. Extensive experiments are conducted on both holistic person re‐identification datasets (Market‐1501, DukeMTMC‐ReID) and occluded person re‐identification datasets (Occluded‐DukeMTMC, Occluded‐ReID). The results demonstrate the superior performance of the authors proposed model, and visualisations of the generation and matching modules are provided to illustrate their effectiveness. Furthermore, an ablation study is conducted to validate the contributions of the proposed framework.

Abstract Semantic segmentation from a three‐dimensional point cloud is vital in autonomous driving, computer vision, and augmented reality. However, current semantic segmentation does not effectively use the point cloud's local geometric features and contextual information, essential for improving segmentation accuracy. A semantic segmentation network that uses local feature fusion and a multilayer attention mechanism is proposed to address these challenges. Specifically, the authors designed a local feature fusion module to encode the geometric and feature information separately, which fully leverages the point cloud's feature perception and geometric structure representation. Furthermore, the authors designed a multilayer attention pooling module consisting of local attention pooling and cascade attention pooling to extract contextual information. Local attention pooling is used to learn local neighbourhood information, and cascade attention pooling captures contextual information from deeper local neighbourhoods. Finally, an enhanced feature representation of important information is obtained by aggregating the features from the two deep attention pooling methods. Extensive experiments on large‐scale point‐cloud datasets Stanford 3D large‐scale indoor spaces and SemanticKITTI indicate that authors network shows excellent advantages over existing representative methods regarding local geometric feature description and global contextual relationships.

Abstract Aircraft coating process has been an important part in manufacturing process of modern aviation products. For coating defect detection, the manual observation with naked eyes is usually utilized, which leads to low production efficiency. In this paper, the authors propose the improved YOLOv5‐OBB with the channel‐spatial attention block (CSAB), feature pyramid non‐local module (FPNM) and structured sparsity slimming criterion (SSSC). The CSAB can pay more attention to effective channel information features from the channel dimension and the target information area from the spatial dimension. The effective non‐local module called FPNM is proposed to further improve the detection accuracy. The authors utilize the oriented bounding boxes (OBB) to reduce redundant background information for coating defect detection. In addition, the SSSC is proposed to achieve network slimming and trade‐off between the efficiency and accuracy. The experimental results on several datasets demonstrate the effectiveness of the authors’ scheme, which achieves superior performance.

Abstract Recognition of partial discharge (PD) patterns is essential for insulation diagnosis of covered conductors in overhead lines. Current research has not sufficiently addressed the complex background noise in real environments, and most detection methods depend primarily on feature engineering or deep learning, suggesting potential for improvement in accuracy and efficiency. This has led the authors to propose a PD pattern recognition algorithm that integrates feature selection and deep learning. This algorithm incorporates the design of a discrete wavelet denoising function specifically tailored to the characteristics of PD for data preprocessing. It employs Bayesian optimization algorithms and light gradient boosting machines for characterizing corona discharge features. Furthermore, it develops multi‐scale clustering features and phase‐resolved features for feature fusion, and constructs insightful features based on the light gradient boosting machine. Finally, a novel deep learning model is formulated, demonstrating exceptional detection performance for early faults in covered conductors. Experimental results show that this algorithm attains an Matthews correlation coefficient score of 0.814, a 13.2% improvement over the baseline algorithm's 0.719, and a speed increase of 39.18%. The final accuracy amounts to 97.85%. This algorithm demonstrates exceptional performance in detecting early insulation faults in conductors.

Abstract Diabetic retinopathy (DR), the main cause of irreversible blindness, is one of the most common complications of diabetes. At present, deep convolutional neural networks have achieved promising performance in automatic DR detection tasks. The convolution operation of methods is a local cross‐correlation operation, whose receptive field determines the size of the local neighbourhood for processing. However, for retinal fundus photographs, there is not only the local information but also long‐distance dependence between the lesion features (e.g. hemorrhages and exudates) scattered throughout the whole image. The proposed method incorporates correlations between long‐range patches into the deep learning framework to improve DR detection. Patch‐wise relationships are used to enhance the local patch features since lesions of DR usually appear as plaques. The Long‐Range unit in the proposed network with a residual structure can be flexibly embedded into other trained networks. Extensive experimental results demonstrate that the proposed approach can achieve higher accuracy than existing state‐of‐the‐art models on Messidor and EyePACS datasets.

Abstract Although low‐rank representation (LRR)‐based subspace learning has been widely applied for feature extraction in computer vision, how to enhance the discriminability of the low‐dimensional features extracted by LRR based subspace learning methods is still a problem that needs to be further investigated. Therefore, this paper proposes a novel low‐rank preserving embedding regression (LRPER) method by integrating LRR, linear regression, and projection learning into a unified framework. In LRPER, LRR can reveal the underlying structure information to strengthen the robustness of projection learning. The robust metric L2,1‐norm is employed to measure the low‐rank reconstruction error and regression loss for moulding the noise and occlusions. An embedding regression is proposed to make full use of the prior information for improving the discriminability of the learned projection. In addition, an alternative iteration algorithm is designed to optimise the proposed model, and the computational complexity of the optimisation algorithm is briefly analysed. The convergence of the optimisation algorithm is theoretically and numerically studied. At last, extensive experiments on four types of image datasets are carried out to demonstrate the effectiveness of LRPER, and the experimental results demonstrate that LRPER performs better than some state‐of‐the‐art feature extraction methods.

Abstract Due to the huge cost of manual annotations, the labelled data may not be sufficient to train a dynamic facial expression (DFR) recogniser with good performance. To address this, the authors propose a multi‐modal pre‐training method with a pseudo‐label guidance mechanism to make full use of unlabelled video data for learning informative representations of facial expressions. First, the authors build a pre‐training dataset of videos with aligned vision and audio modals. Second, the vision and audio feature encoders are trained through an instance discrimination strategy and a cross‐modal alignment strategy on the pre‐training data. Third, the vision feature encoder is extended as a dynamic expression recogniser and is fine‐tuned on the labelled training data. Fourth, the fine‐tuned expression recogniser is adopted to predict pseudo‐labels for the pre‐training data, and then start a new pre‐training phase with the guidance of pseudo‐labels to alleviate the long‐tail distribution problem and the instance‐class confliction. Fifth, since the representations learnt with the guidance of pseudo‐labels are more informative, a new fine‐tuning phase is added to further boost the generalisation performance on the DFR recognition task. Experimental results on the Dynamic Facial Expression in the Wild dataset demonstrate the superiority of the proposed method.

Abstract License plate detection is an important task in Intelligent Transportation Systems (ITS) and has a wide range of applications in vehicle management, traffic control, and public safety. In order to improve the accuracy and speed of mobile recognition, an improved lightweight YOLOv5s model is proposed for license plate detection. First, an improved Stemblock network is used to replace the original Focus layer in the network, which ensures strong feature expression capability and reduces a large number of parameters to lower the computational complexity; then, an improved lightweight network, ShuffleNetv2, is used to replace the backbone network of the YOLOv5s, which makes the model lighter and ensures the detection accuracy at the same time. Then, a feature enhancement module is designed to reduce the information loss caused by the rearrangement of the backbone network channels, which facilitates the information interaction in the feature fusion process; finally, the low‐, medium‐ and high‐level features in the Shufflenetv2 network structure are fused to form the final high‐level output features. Experimental results on the CCPD dataset show that compared to other methods this paper obtains better performance and faster speed in the license plate detection task, in which the average precision mean value reaches 96.6%, and can achieve a detection speed of 43.86 frame/s, and the parameter volume is reduced to 5.07 M.

Abstract Anticoagulation is vital to maintain blood fluidic status and physiological functions in the field of clinical blood‐related procedures. Here, novel biomimetic anticoagulated porous inverse opal hydrogel particles is presented as anticoagulant bearing dynamic screening capability. The inverse opal hydrogel particles possess abundant sulfonic and carboxyl groups, which serve as binding sites with multiple coagulation factors and inhibit the blood coagulation process. Owing to the variations of refractive index and pore sizes during the binding process, the particles appeared corresponding structure color variations, which can be adopted as sensory index of anticoagulation. Based on these features, a sensor containing these diverse structure color particle units is constructed for pattern recognition of coagulation factors level in clinical plasma samples. By analyzing the sensory information of the unit, the colorimetric “fingerprint” for each target can be obtained and summarized as a database. Besides, a portable test‐strip integrating sensory units is developed to distinguish the sample regarding abnormal coagulation factors‐derived diseases via multivariate data analysis. It is believed that such biomimetic anticoagulated structural color particles and their derived sensor will open new avenue for clinical detection and disease diagnosis.

Abstract Considering the defect of object recognition with single global feature or local feature, in this letter the authors propose a shape interest points descriptor (SIPD) for object recognition. Particularly, the authors extract the shape features by the improved HU moments and then interest points by Speeded up Robust Feature (SURF). Object recognition is carried out by similarity measure. Because the influence of the improved HU moments and SURF for object recognition is different, the two different similarity measures are fused effectively by using different weight factors. Experimental results show that the author’ proposed method is effective and robust to the changes of scale, viewing angle, illumination and noise when compared with the other representative methods.

Abstract Background Lung ultrasound (LUS) is a noninvasive tool for examining respiratory distress patients. The lung ultrasound score (LUSS) can be used to quantify and monitor lung aeration loss with good reliability. Hypothesis/Objectives Assess the reliability of a new LUSS among raters with different levels of experience and determine how well the same raters agree on identifying patterns of LUS abnormalities. Animals Forty LUS examinations of dogs and cats and 320 videos were reviewed from a digital database. Methods Retrospective reliability study with post hoc analysis. Protocolized LUS were randomly selected; intrarater and interrater reliability of the LUSS and pattern recognition agreement among 4 raters with different levels of experience in LUS were tested. Results The intrarater intraclass correlation coefficient (ICC) single measurement, absolute agreement, and 2‐way mixed effects model was 0.967 for the high‐experience rater (H‐Exp), 0.963 and 0.952 for the medium‐experience raters (M‐Exp‐1; M‐Exp‐2), and 0.950 for the low‐experience rater (L‐Exp). The interrater ICC average measurement, absolute agreement, and 2‐way random effects model among the observers was 0.980. The Fleiss' kappa (k) values showed almost perfect agreement (k = 1) among raters in identifying pleural effusion and translobar tissue‐like pattern, strong agreement for A‐lines (k = 0.881) and B‐lines (k = 0.806), moderate agreement (k = 0.693) for subpleural loss of aeration, and weak agreement (k = 0.474) for irregularities of the pleural line. Conclusions and Clinical Importance Our results indicate excellent intra‐ and interrater reliability for LUS scoring and pattern identification, providing a foundation for the use of the LUSS in emergency medicine and intensive care.

Abstract Peak‐valley period partition of load curve is a key aspect of time‐of‐use (ToU) tariff to improve power load characteristics, such as shifting peak loads towards valley time periods. Fuzzy clustering algorithm is an effective and popular method commonly used to solve the peak‐valley period partition of load curves, but it still encounters the difficulty of dividing some data within the boundary regions of different time periods. Therefore, this paper presents a peak‐valley period partition and abnormal time correction scheme for ToU tariffs under typical daily load curves based on improved fuzzy C‐means (FCM) clustering algorithm. In order to improve the accuracy of peak‐valley period partition, modified fuzzy membership functions are proposed to improve the initialization of FCM clustering, and a loss function‐based method is presented for calculating the fuzzy parameters of those membership functions. To resolve the problem of abnormal time partitioning within the boundaries of different time periods, an abnormal time period recognition model and a correction model based on fuzzy subsethood are proposed to obtain the final corrected peak‐valley time period partitioning results. On the MATLAB R2020b platform, the effectiveness of the proposed method is verified through two real daily load curves with a time resolution of 5 min.

Abstract Heatmap‐based regression (HBR) methods have dominated for a long time in the face alignment field while these methods need complex design and post‐processing. In this study, the authors propose an end‐to‐end and simple enough coordinate‐based regression (CBR) method called Dynamic Deformable Transformer (DDT) for face alignment. Unlike general pre‐defined landmark queries, DDT uses Dynamic Landmark Queries (DLQs) to query landmarks' classes and coordinates together. Besides, DDT adopts a deformable attention mechanism rather than a regular attention mechanism which has a faster convergence speed and lower computational complexity. Experiment results on three mainstream datasets 300W, WFLW, and COFW demonstrate DDT exceeds the state‐of‐the‐art CBR methods by a large margin and is comparable to the current state‐of‐the‐art HBR methods with much less computational complexity.

Abstract Most of the successful person re‐ID models conduct supervised training and need a large number of training data. These models fail to generalise well on unseen unlabelled testing sets. The authors aim to learn a generalisable person re‐identification model. The model uses one labelled source dataset and one unlabelled target dataset during training and generalises well on the target testing set. To this end, after a feature extraction by the ResNext‐50 network, the authors optimise the model by three loss functions. (a) One loss function is designed to learn the features of the target domain by tuning the distances between target images. Therefore, the trained model will be more robust to overcome the intra‐domain variations in the target domain and generalises well on the target testing set. (b) One triplet loss is used which considers both source and target domains and makes the model learn the inter‐domain variations between source and target domain as well as the variations in the target domain. (c) Also, one loss function is for supervised learning on the labelled source domain. Extensive experiments on Market1501 and DukeMTMC re‐ID show that the model achieves a very competitive performance compared with state‐of‐the‐art models and also it requires an acceptable amount of GPU RAM compared to other successful models.

Abstract Bird's nest defect is an important cause of transmission line faults. To achieve accurate detection of bird nest defects in complex scenarios, a bird nest defect detection model for transmission lines was proposed that combines domain knowledge and occlusion reasoning networks. On the one hand, the model utilized the domain knowledge of the location of the bird's nest, using edge detection to obtain tower area information to constrain the location of candidate frames. This helps to reduce the false detection caused by complex backgrounds. On the other hand, on the basis of analyzing the occlusion characteristics of bird nests, the model employed occlusion reasoning networks that randomly erase features at the feature level to simulate the occlusion of bird nests in real scenes and improve the model's detection capability for occluded targets. Additionally, a multi‐scale feature fusion algorithm was designed in this paper to adapt the model to the scale variations of bird nests in aerial images. Experimental results demonstrate that the model outperforms advanced target detection models and other bird nest defect detection methods, with an AP50 of 78.8% and an AR10 of 72.4% for defect detection.

Abstract This study enhances the traditional sparrow search algorithm (SSA) to address its weaknesses, such as poor convergence rates and precision due to a tendency to fall into local optima. The improved version, OCSSA, integrates the Osprey optimization algorithm (OOA) and Cauchy mutation. Logistic chaotic mapping is used for initial population generation to increase genetic diversity. OOA boosts global exploration in the producer phase, and Cauchy mutation in the scroungers phase disrupts suboptimal solutions, enhancing the algorithm's ability to avoid local optima. OCSSA's performance, validated through ten benchmark functions and fault diagnosis optimization tasks, significantly improves convergence speed and accuracy, proving its effectiveness in complex optimization challenges.

Abstract Data hallucination generates additional training examples for novel classes to alleviate the data scarcity problem in few‐shot learning (FSL). Existing hallucination‐based FSL methods normally train a general embedding model first by applying information extracted from base classes that have abundant data. In those methods, hallucinators are then built upon the trained embedding model to generate data for novel classes. However, these hallucination methods usually rely on general‐purpose embeddings, limiting their ability to generate task‐oriented samples for novel classes. Recent studies have shown that task‐specific embedding models, which are adapted to novel tasks, can achieve better classification performance. To improve the performance of example hallucination for tasks, a task‐oriented embedding model is used in the proposed method to perform task‐oriented generation. After the initialization, the hallucinator is finetuned by applying a task‐oriented embedding model with the guidance of a teacher–student mechanism. The proposed task‐oriented hallucination method contains two steps. An initial embedding network and an initial hallucinator are trained with a base dataset in the first step. The second step contains a pseudo‐labelling process where the base dataset is pseudo‐labelled using support data of the few‐shot task and a task‐oriented fine‐tuning process where the embedding network and hallucinator are adjusted simultaneously. Both the embedding network and the hallucinator are updated with the support set and the pseudo‐labelled base dataset using knowledge distillation. The experiments are conducted on four popular few‐shot datasets. The results demonstrate that the proposed approach outperforms state‐of‐the‐art methods with 0.8% to 4.08% increases in classification accuracy for 5‐way 5‐shot tasks. It also achieves comparable accuracy to state‐of‐the‐art methods for 5‐way 1‐shot tasks.

Abstract With the exponential growth of video data, action recognition has become an increasingly important area of study. Despite various advancements, achieving a balance between detection accuracy and lightness remains a formidable challenge, primarily due to the complexity of existing action recognition models. To address this issue, DenseGCN is developed, a lightweight network designed to optimize accuracy and efficiency. The aim was to create a detection model that has high accuracy while remaining lightweight for real‐world applications. DenseGCN operates via a unique three‐level feature fusion system. The initial stage involves the Multi‐level Fusion Network (MlFN), which contains dense connections and a Spatial‐Temporal Fusion Attention module (STF‐Att), designed to eliminate bias in feature extraction caused by deep networks. In the next stage, RefineBone tackles optimization issues in low‐dimensional feature layers by leveraging high‐dimensional feature layers, thus avoiding gradient stacking. Finally, the Multi‐temporal Fusion Feature Pyramid Network (MF‐FPN) generates a discriminative classification feature map by repetitively combining data from multiple dimensions. This strategy has proven successful in refining the extracted feature, allowing for discriminative feature extraction even with a reduced number of channels. This efficient design not only contributes to further research in developing lightweight networks but also offers enhanced possibilities for real‐world implementations. In two large‐scale datasets, NTU RGB+D 60 and 120, DenseGCN outperformed other state‐of‐the‐art methods, achieving an accuracy of 92.7% on the X‐View benchmark of the NTU RGB+D 60 dataset. The DenseGCN is 10.2 × faster and 10 × smaller than the spatial temporal graph attention network (STGAT) proposed in 2022 while retaining very competitive accuracy. The findings suggest that this model significantly improves the quality of feature extraction. As a result, DenseGCN presents a remarkable balance between accuracy and lightness.

Abstract Skeleton‐based neural networks have been considered a focus for human action recognition (HAR). It is noteworthy that the existing skeleton‐based methods are not capable of combining the spatial and temporal features reasonably to derive more effective high‐level representations, and it continues to be a challenging task of learning and representing the skeleton action discriminatively. In this study, a novel two‐stream spatiotemporal network (TSTN) is proposed, which is capable of processing the spatial and temporal features respectively and collectively to achieve a better representation and understanding of human action. The temporal branch stacks three gate recurrent unit (GRU) blocks in a new architecture to encode the temporal correlations from different aspects of human action, achieving high‐level temporal semantic feature expressions. The spatial branch encodes the spatial features with multi‐stacked graph convolutional network (GCN) blocks. Self‐attention mechanisms incorporated with the graph structure of the skeleton are explored to add weight influence and structural hints to further enhance the performance. The experimental results verify the effectiveness and superiority of the proposed model in skeleton action recognition; the model reaches state‐of‐the‐art on specific datasets.

Abstract Human keypoints detection is different from general detection tasks and requires networks that can learn visual information and anatomical constraints. Since CNN is excellent in extracting texture features of images and transformer can learn the correlation among keypoints well, many CTPNets (CNN+transformer type human pose estimation networks) have emerged. However, these networks are unconcerned with the processing of the features extracted from the CNN and naturally expand only from the channel dimension, ignoring the spatial features in the visual information that are essential for complex detection tasks like pose estimation. So the channel spatial integrated transformer for human pose estimation, termed CSIT, is proposed. The visual information are summarized as texture and spatial information, and a parallel network is used to expand the feature maps in the channel and spatial dimensions to learn texture features and spatial features respectively. In addition, anatomically constrained information is learned by keypoint embeddings. At the end of the network, the 1D vector representation method with more advanced performance and more compatible with transformer's characteristics is used to predict keypoints. Experiments show that CSIT outperforms the mainstream CTPNets on the COCO test‐dev dataset, and also show satisfactory results on the MPII dataset.

Abstract Facial expression recognition (FER) is widely used in many fields. To further improve the accuracy of FER, this paper proposes a method based on double‐code LBP‐layer spatial‐attention network (DLSANet). The backbone model for the DLSANet is an emotion network (ENet), which is modified with a double‐code LBP (DLBP) layer and a spatial attention module. The DLBP layer is at the front of the first convolutional layer. More valuable features can be extracted by inputting the image processed by DLBP into convolutional layers. The JAFFE and CK+ datasets are used, which contain seven expressions: happiness, anger, disgust, neutral, fear, sadness, and surprise. The average of fivefold cross‐validation shows that DLSANet achieves a recognition accuracy of 93.81% and 98.68% on the JAFFE and CK+ datasets. The experiment reveals that the DLSANet can produce better classification results than state‐of‐the‐art methods.

Abstract There are many defects in gas‐insulated metal‐enclosed switchgear (GIS) and gas‐insulated transmission lines (GIL) that may cause accidents, such as floating potentials, metal particles, air gaps, and cracks in support insulators. Nevertheless, to‐this‐date, there is no effective method to recognize them to adjust their operation status. Accordingly, this study compared three typical defect models to explore the partial discharge process, and three typical experimental models were established to identify the characteristic parameters of acoustic and electrical information in the development of defects. The research study shows that that as the voltage increases, the partial discharge of free‐metal‐particle defects and air‐gap defects obtained by the pulse‐current method increase gradually. A sudden increase also occurred in the partial discharge of the floating potential defect. The discharge amount of free conductive particles obtained by an ultrahigh‐frequency detection method was distributed symmetrically on both sides of the applied voltage peak, and the other two defects were on one side of the peak and yielded an obvious phase difference. The particle collision signal obtained by the ultrasonic method was obvious, but had no obvious phase relationship with the applied voltage. However, the obtained floating potential defect information had obvious phase differences, and the ultrasonic method was not sensitive to air‐gap defects. The three typical defects can be identified by the combined method more accurately; this provides a theoretical basis and data support for GIS and GIS voltage tests, detection technologies, and online monitoring methods.

Perceiving surface characteristics through tactile interaction typically requires high‐resolution devices or precise spatial scanning to record and analyze a significant amount of information. However, most available tactile sensors require complicated technological processes, redundant layouts, and data acquisition circuits, which limits their ability to provide a real‐time static perception and feedback for potential applications such as robotic manipulation. Drawing inspiration from the sliding tactile (ST) perception mode of the human fingertip, a robust and flexible ST sensor with a low array density of 2.7 cells cm−2 is reported. This innovative sensor has a soft and cambered configuration that allows it to rapidly and accurately recognize the 3D surface features of objects, including grooves as small as 500 μm. Benefiting from the strong correlation between collected electronic responding and local deformation of sensing cell, the ST sensor can adaptively reconstruct surface patterns with the assistance of deep learning, even on unstructured objects. The pattern recognition system based on the robot is demonstrated by accurately classifying a set of mahjong tiles with nearly 100% accuracy, surpassing human tactile perception capabilities in the same task.

Abstract Aerial targets grouping is an important content of air battlefield situation awareness, and intelligent aerial targets grouping is an urgent need for intelligent air combat. Traditional aerial targets grouping methods have problems such as difficulty extracting air target features, difficulty designing formation similarity measures, and lack of data samples. To solve the above problems, this paper proposes a relational network based aerial targets grouping method to achieve end‐to‐end grouping under small sample conditions. Firstly, an adaptive clustering algorithm is proposed to preliminarily cluster the aerial targets in the air battlefield situation to obtain the group to be tested. Then, a representation of aerial target groups is designed, which unifies the multi‐dimensional features of the group's temporal, spatial, and attribute. Next, based on this representation, combining 3D convolutional neural networks (3DCNN) and long short‐term memory (LSTM), a neural network structure that can automatically extract multi‐dimensional features of groups is designed. After that, the relation module is used to learn the distance metric between groups. After training the relation network, the type of group to be tested can be identified, thereby achieving the goal of grouping. A dataset of 30 types of aerial target group patterns was established through simulation. The ablation experiments were conducted to verify the performance of each module, and the experimental results showed that the designed feature extraction module can effectively extract temporal sample features, while the designed relationship module performs better in extracting temporal sample information. By comparing several typical traditional aerial target grouping methods, the proposed method improves the grouping accuracy by 11.2% to 13.4%, achieving an accuracy of 91.1% in the 5‐way 1‐shot task and 95.6% in the 5‐way 5‐shot task.

Abstract The authors propose encoding and decoding network‐based normalizing flows (EaD‐Flow) for the detection and localization of anomalies in an image. Multiple networks of the EaD‐Flow estimate probability densities of tenor image. The anomaly localization model based on multiple networks shows superior anomaly detection and localization performance.

Abstract This brief proposes a field programmable gate array (FPGA) based real‐time inference machine for A‐mode ultrasonic echo pattern recognition. The proposed scheme can aid a probe positioning of single‐transducer based ultrasound devices by detecting specific echo patterns on a scanline. The inference machine is based on a combination of a support vector machine and a finite state machine. The proposed inference machine utilizes a minimal pre‐processing for a feature extraction, considering a nature of distinctive echo patterns. In addition, a primary computation of linear support vector machine is performed sequentially, and the finite state machine for robust binary classification is relatively compact. As a result, the overall structure of the inference machine can be concisely implemented on FPGA. The proposed inference machine was assembled with a customized A‐mode ultrasound scanner. The maximum utilization percentage of look‐up‐tables in FPGA was less than 1.5%. In evaluation of inferences, the resultant sensitivity and the specificity were 96.2% and 99.87%, respectively.

Abstract Low‐temperature composite insulation is commonly applied in high‐temperature superconducting apparatus while partial discharge (PD) is found to be an important indicator to reveal insulation statues. In order to extract feature parameters of PD signals more effectively, a method combined variational mode decomposition with multi‐scale entropy and image feature is proposed. Based on the simulated test platform, original and noisy signals of three typical PD defects were obtained and decomposed. Accordingly, relative moments and grayscale co‐occurrence matrix were employed for feature extraction by K‐modal component diagram. Afterwards, new PD feature vectors were obtained by dimension reduction. Finally, effectiveness of different feature extraction methods was evaluated by pattern recognition based on support vector machine and K‐nearest neighbour. Result shows that the proposed feature extraction method has a higher recognition rate by comparison and is robust in processing noisy signals.

Abstract Arecanut disease identification is a challenging problem in the field of image processing. In this work, we present a new combination of multi‐gradient‐direction and deep convolutional neural networks for arecanut disease identification, namely, rot, split and rot‐split. Due to the effect of the disease, there are chances of losing vital details in the images. To enhance the fine details in the images affected by diseases, we explore multi‐Sobel directional masks for convolving with the input image, which results in enhanced images. The proposed method extracts arecanut as foreground from the enhanced images using Otsu thresholding. Further, the features are extracted for foreground information for disease identification by exploring the ResNet architecture. The advantage of the proposed approach is that it identifies the diseased images from the healthy arecanut images. Experimental results on the dataset of four classes (healthy, rot, split and rot‐split) show that the proposed model is superior in terms of classification rate.

Abstract The Multiple Classifier System (or classifier ensemble) is the consensus of different clustering algorithms that can provide high accuracy for the best partition and thus overcome the constraints of conventional approaches based on single classifiers. The MCS is divided into two stages: Partition creation and partition combining. The potential benefits of this methodology in unsupervised land cover categorization utilizing synthetic, composite, and remotely sensed data are investigated in this paper. Four clustering algorithms are used for the MCS's first step, and according to the WB index, the best‐unsupervised classification is obtained. In the second stage, relabeling and, voting approaches are then applied. The MCS's experimental results outperform the individual clustering outcomes in terms of accuracy.

Abstract The Gaofen‐7 (GF‐7) satellite is China's first submeter high‐resolution optical stereo‐mapping satellite. The satellite is equipped with two area‐array laser footprint cameras to capture the laser spots emitted by laser altimeters. To accurately identify and locate the geographical positions of laser spots, a stepwise geometric calibration method with ground laboratory information for high‐resolution area‐array camera is proposed in this paper. It includes the internal and external calibration. First, the initial values of the camera's internal parameters are determined based on the laboratory calibration results. Then, based on a high‐precision reference ortho‐image and a digital elevation model, the ground control points (GCPs) for multi‐scene footprint images are obtained by image matching. Finally, the external and internal calibration methods for the laser footprint camera are implemented successively. The experimental results indicate that this method can restore geometric imaging relationships with high precision. After the internal and external calibration, the systematic residual error of the image point is eliminated, the remaining residual achieves 0.38 pixel, and the accuracy of direct positioning on the ground is better than 5.0 m, and the elevation accuracy is within 1.5 m.

The diagnosis and prevention of Alzheimer's disease plays an important role in improving patient cognition. It can be seen from the current situation that the diagnosis of Alzheimer's disease is still poor because it is affected by many factors. Based on this, combined with the symptoms of Alzheimer's disease, this study used computer‐aided to diagnose the symptoms of patients. First of all, this study analyses classical machine learning and chooses the appropriate model for diagnosis. Next, this study constructs a diagnostic system based on a mixed Gaussian model and uses a mixed Gaussian model to predict the probability of different distribution methods. Finally, this study designs experiments to analyse the performance of diagnostic models. Studies have shown that the mixed Gaussian model has a good effect on the automatic diagnosis of Alzheimer's disease, and can provide a theoretical reference for subsequent related research.

Fuzzy c‐means algorithms have been widely utilised in several areas such as image segmentation, pattern recognition and data mining. However, the related studies showed the limitations in facing imbalanced datasets. The maximum fuzzy boundary tends to be located on the largest cluster which is not desirable. The overall fuzzy partition results in false grouping of edge objects and weakens the compactness of cluster. It is important the clusters are delineated by the maximum fuzzy boundary. In this study, a semi‐fuzzy c‐means algorithm is proposed by combining hard partition and soft partition. This study aims to provide an effective partition for the edge objects, such that the compactness of cluster can be improved. The proposed algorithm integrates the semi‐fuzzy c‐means method with the size‐insensitive integrity‐based fuzzy c‐means algorithm. In particular, the latter algorithm has the ability to deal with imbalanced data. With the experiment validation, the proposed algorithm is robust and outperforms the two component algorithms by using synthetic and widely known benchmark datasets.

Colour images are increasingly used in the fields of computer vision, pattern recognition and machine learning, since they can provide more identifiable information than greyscale images. Thus, colour face recognition has attracted accumulating attention. Its key problem is how to remove the similarity between colour component images and take full advantage of colour difference information. Decision‐level similarity reduction between colour component images directly affects the recognition effect, but it has been found in no work. In this study, the authors propose a novel colour face recognition approach named semi‐supervised uncorrelated dictionary learning (SUDL), which realises decision‐level similarity reduction and fusion of all colour components in face images. SUDL employs the labelled and unlabelled colour face image samples into structured dictionary learning to achieve three uncorrelated discriminating dictionaries corresponding to three colour components of face images, and then uses these dictionaries and the sparse coding technique to make a classification decision. Experimental results in multiple public colour face image databases demonstrate that the dictionary decorrelation, structured dictionary learning and unlabelled samples used in the proposed approach are effective and reasonable, and the proposed approach outperforms several representative colour face recognition methods in recognition rates, despite of its poor time performance.

Abstract In the present paper a data‐driven hard cluster analysis derived from a novel similarity measure is proposed to support financial investors in their portfolio management decision‐making process. The main objective of the proposed method is to provide a less arbitrary learning procedure to quantify similarity levels between investment alternatives (pairwise) as well as revealing clustering patterns (whole sample). This is especially useful during periods of high volatility, when investment alternatives tend to become more similar and, therefore, harder to distinguish between themselves. The method dynamics may be readily interpreted through a clear data visualisation. The advantages and caveats of the proposed method is compared to the most popular class of cluster analysis, applied to the well‐known Fisher's Iris dataset. Such results show a slightly superior performance of the proposed method but, most importantly, through remarkably different clustering allocation approaches. Moreover, further empirical results applied to daily data reflecting a period of 15 years of time series of economies/stock markets of the Group of Seven (G7) illustrate the potential practical usefulness of the proposed unsupervised learning method, particularly, for portfolio strategy, asset allocation, and investment diversification.

The aim of the current study was to compare the performance of fully automated software with human expert interpretation of single-voxel proton magnetic resonance spectroscopy (1H-MRS) spectra in the assessment of breast lesions. Breast magnetic resonance imaging (MRI) (including contrast-enhanced T1-weighted, T2-weighted, and diffusion-weighted imaging) and 1H-MRS images of 74 consecutive patients were acquired on a 3-T positron emission tomography-MRI scanner then automatically imported into and analyzed by SpecTec-ULR 1.1 software (LifeTec Solutions GmbH). All ensuing 117 spectra were additionally independently analyzed and interpreted by two blinded radiologists. Histopathology of at least 24 months of imaging follow-up served as the reference standard. Nonparametric Spearman's correlation coefficients for all measured parameters (signal-to-noise ratio [SNR] and integral of total choline [tCho]), Passing and Bablok regression, and receiver operating characteristic analysis, were calculated to assess test diagnostic performance, as well as to compare automated with manual reading. Based on 117 spectra of 74 patients, the area under the curve for tCho SNR and integrals ranged from 0.768 to 0.814 and from 0.721 to 0.784 to distinguish benign from malignant tissue, respectively. Neither method displayed significant differences between measurements (automated vs. human expert readers, p > 0.05), in line with the results from the univariate Spearman's rank correlation coefficients, as well as the Passing and Bablok regression analysis. It was concluded that this pilot study demonstrates that 1H-MRS data from breast MRI can be automatically exported and interpreted by SpecTec-ULR 1.1 software. The diagnostic performance of this software was not inferior to human expert readers., (© 2023 The Authors. NMR in Biomedicine published by John Wiley & Sons Ltd.)

Complex neutrosophic set (CNS) is a modified version of the complex fuzzy set, to cope with complicated and inconsistent information in the environment of fuzzy set theory. The CNS is characterised by three functions expressing the degree of complex-valued membership, complex-valued abstinence and degree of complex-valued non-membership. The aim of this manuscript is to initiate the novel dice similarity measures and generalised dice similarity using CNS. The special cases of the investigated measures are discussed with the help of some remarks. Moreover, some distance measures based on CNS are also proposed in this manuscript. Then, the authors applied the generalised dice similarity measures and weighted generalised dice similarity measures using CNS to the pattern recognition model to examine the reliability and superiority of the established approaches. The advantages and comparative analysis of the proposed measures with existing measures are also discussed in detail. At last, a numerical example is provided to illustrate the validity and applicability of the presented measures.

The feature extraction and pattern recognition techniques are of great importance to assess the insulation condition of gas-insulated switchgear. In this work, the ultra-high-frequency partial discharge (PD) signals generated from four types of typical insulation defects are analysed using S-transform, and the greyscale image in time-frequency representation is divided into five regions according to the cutoff frequencies of TE(m)(1) modes. Then, the three low-order moments of every subregion are extracted and the feature selection is performed based on the J criterion. To confirm the effectiveness of selected moment features after considering the electromagnetic modes, the support vector machine, k-nearest neighbour and particle swarm-optimised extreme learning machine (ELM) are utilised to classify the type of PD, and they achieve the recognition accuracies of 92, 88.5 and 95%, respectively. In addition, the results show that the ELM offers good generalisation performance at the fastest learning and testing speeds, thus more suitable for a real-time PD detection.

The measurement of the right and wrong results of cognitive tasks has important applications in many commercial and educational areas such as the drivers’ training system, the simulation training and online learning system. This Letter aims to distinguish the heartbeat pattern of cognitively wrong responses to that of cognitively right responses based on the electrocardiogram (ECG) through 36 subjects with different professional backgrounds. The experimental design methods were double-digit and five-digit addition/subtraction, which were blindly selected by subjects from a black box. Through the R–R interval (RRI) series obtained from the ECG data, some linear, nonlinear and moment features were extracted to evaluate the cognitive task results by using pattern recognition methods. The binary classification of RRI datasets indicated that autonomic nerve patterns of the right and wrong cognitive heartbeat responses were distinguishable.

Gear and its transmission are widely used in different transmission systems, and its complicated and changeable condition brings a series of problems to the fault feature extraction and diagnosis. In recent years, deep learning techniques have been gradually applied to feature extraction and pattern recognition, and the features of feature extraction and fault diagnosis in complex working environments have shown certain advantages. This study is based on stacked autoencoder under deep learning model, and improve training network performance by modified activation function. Through the network training before and after the experiment done, and to extract the fault feature data comparison in testing, improving network after activation function to extract fault features showed a greater advantage, can be a very good application in practical fault feature extraction.

Health care networks for Parkinson's disease (PD) already exist and have been already proposed in the literature, but most of them are not able to analyse the vast volume of data generated from medical examinations and collected and organised in a pre-defined manner. In this work, the authors propose a novel health care network based on big data analytics for PD. The main goal of the proposed architecture is to support clinicians in the objective assessment of the typical PD motor issues and alterations. The proposed health care network has the ability to retrieve a vast volume of acquired heterogeneous data from a Data warehouse and train an ensemble SVM to classify and rate the motor severity of a PD patient. Once the network is trained, it will be able to analyse the data collected during motor examinations of a PD patient and generate a diagnostic report on the basis of the previously acquired knowledge. Such a diagnostic report represents a tool both to monitor the follow up of the disease for each patient and give robust advice about the severity of the disease to clinicians.