Generalized Features Research Articles

Few Shot Object Detection for SAR Images via Feature Enhancement and Dynamic Relationship Modeling

Current Synthetic Aperture Radar (SAR) image object detection methods require huge amounts of annotated data and can only detect the categories that appears in the training set. Due to the lack of training samples in the real applications, the performance decreases sharply on rare categories, which largely inhibits the detection model from reaching robustness. To tackle this problem, a novel few-shot SAR object detection framework is proposed, which is built upon the meta-learning architecture and aims at detecting objects of unseen classes given only a few annotated examples. Observing the quality of support features determines the performance of the few-shot object detection task, we propose an attention mechanism to highlight class-specific features while softening the irrelevant background information. Considering the variation between different support images, we also employ a support-guided module to enhance query features, thus generating high-qualified proposals more relevant to support images. To further exploit the relevance between support and query images, which is ignored in single class representation, a dynamic relationship learning paradigm is designed via constructing a graph convolutional network and imposing orthogonality constraint in hidden feature space, which both make features from the same category more closer and those from different classes more separable. Comprehensive experiments have been completed on the self-constructed SAR multi-class object detection dataset, which demonstrate the effectiveness of our few-shot object detection framework in learning more generalized features to both enhance the performance on novel classes and maintain the performance on base classes.

TransMEF: A Transformer-Based Multi-Exposure Image Fusion Framework Using Self-Supervised Multi-Task Learning

In this paper, we propose TransMEF, a transformer-based multi-exposure image fusion framework that uses self-supervised multi-task learning. The framework is based on an encoder-decoder network, which can be trained on large natural image datasets and does not require ground truth fusion images. We design three self-supervised reconstruction tasks according to the characteristics of multi-exposure images and conduct these tasks simultaneously using multi-task learning; through this process, the network can learn the characteristics of multi-exposure images and extract more generalized features. In addition, to compensate for the defect in establishing long-range dependencies in CNN-based architectures, we design an encoder that combines a CNN module with a transformer module. This combination enables the network to focus on both local and global information. We evaluated our method and compared it to 11 competitive traditional and deep learning-based methods on the latest released multi-exposure image fusion benchmark dataset, and our method achieved the best performance in both subjective and objective evaluations. Code will be available at https://github.com/miccaiif/TransMEF.

Performance Evaluation Method of Library Knowledge Management Based on Data Mining

With the development of information technology and computer technology, the work of the library is becoming more and more digitized and networked. This article mainly studies the performance evaluation method of library knowledge management based on data mining. This paper uses attribute-oriented induction algorithm to mine generalized features. First, scan the entire data set to obtain different values of all attributes. Statistics, classification, and analysis of historical records help to understand the usage of books and periodicals and perform predictive analysis. This article uses statistical weighted weight calculation formula to calculate the library knowledge management ability evaluation index weight. The evaluation of the knowledge management ability of the library mainly adopts the questionnaire survey method, and the knowledge management status of the library is deeply understood in the form of interviews on the spot, and the obtained evaluation data and materials of the knowledge management ability of the library are organized and statistics. In order to prevent the model from remembering the patterns of the training set too deeply, to make the model more general, and to adapt to unknown data well, we use the test set to rest the model. The part of the data set that has not been used in the process of modeling and testing correction can be used to estimate the effect of the model, or to compare the effect of the model. For expert value, the value of the office is 1.02, which is approximately equal to 1, which means that the value and cost of the office are basically equal in the knowledge service of the library. The results show that the combination of knowledge management and university library management will help to form a systematic theoretical framework and behavioral model framework in the research of knowledge management models and strategies in university libraries.

Internal Defects Detection Method of the Railway Track Based on Generalization Features Cluster Under Ultrasonic Images

There may be several internal defects in railway track work that have different shapes and distribution rules, and these defects affect the safety of high-speed trains. Establishing reliable detection models and methods for these internal defects remains a challenging task. To address this challenge, in this study, an intelligent detection method based on a generalization feature cluster is proposed for internal defects of railway tracks. First, the defects are classified and counted according to their shape and location features. Then, generalized features of the internal defects are extracted and formulated based on the maximum difference between different types of defects and the maximum tolerance among same defects’ types. Finally, the extracted generalized features are expressed by function constraints, and formulated as generalization feature clusters to classify and identify internal defects in the railway track. Furthermore, to improve the detection reliability and speed, a reduced-dimension method of the generalization feature clusters is presented in this paper. Based on this reduced-dimension feature and strongly constrained generalized features, the K-means clustering algorithm is developed for defect clustering, and good clustering results are achieved. Regarding the defects in the rail head region, the clustering accuracy is over 95%, and the Davies-Bouldin index (DBI) index is negligible, which indicates the validation of the proposed generalization features with strong constraints. Experimental results prove that the accuracy of the proposed method based on generalization feature clusters is up to 97.55%, and the average detection time is 0.12 s/frame, which indicates that it performs well in adaptability, high accuracy, and detection speed under complex working environments. The proposed algorithm can effectively detect internal defects in railway tracks using an established generalization feature cluster model.

Generalized absence seizures: Where do we stand today?

AbstractGeneralized absence seizures are discussed here with respect to changing scientific concepts and newer findings, challenging the strict dichotomy of “generalized” vs. “focal” which may be counterproductive to the development of new ideas and treatments. Data from semiology, structural, and functional studies using quantitative electroencephalography (EEG), video-EEG monitoring, magnetoencephalography, magnetic resonance imaging, and positron emission tomography as well as neuropathology suggest a nosological spectrum from focal to generalized seizure-generating mechanisms. The terms “focal” and “generalized” as used to describe seizures represent abstractions and idealizations rather than ontological entities. The new terminology and classification of the ILAE (2017), despite its enormous merits, unfortunately also facilitates and fosters reductionist thinking, which may be pragmatic from a utilitarian perspective, but risks hindering scientific developments for the future. For instance, the seizure semiology of absences as a paradigmatic generalized seizure type is now classified as “non-motor seizure.” A detailed phenomenological analysis reveals that motor activity is an integral part of this seizure type. Many motor and so-called automatic absence symptoms can be understood as correlated elements of a general organization plan of the cortex. Corresponding functional and structural changes seen in imaging and high-resolution electrophysiology studies indicate focal contributions to seizure generation and a combination of focal and generalized features. Functional findings indicate that the frontal lobe with its projections to other brain areas may play an important role in generalized absence seizures. In this critical review we assess the evolution of pathophysiological concepts through the ages and present new research approaches for diagnosis and treatment.

Detection of GAN-Generated Images by Estimating Artifact Similarity

Recently, researchers have been dedicated to discovering Generative Adversarial Network (GAN) artifacts and using them to identify generated images. However, current approaches exhibit restricted performance when testing against unseen GAN models, which is also known as the cross-domain scenario. To overcome this limitation, we propose a novel GAN-generated image detection framework by estimating artifact similarity, which is inspired by relation network. The proposed method consists of two stages, including representation learning and representation comparison. For representation learning, ResNet-50 equipped with Instance Normalization in the Shallow layers (ResNet-INS) is constructed as the embedding network to extract generalized features. For representation comparison, Category and Domain-Aware loss function (CDA loss) is designed by leveraging both category and domain information efficiently, which can enlarge inter-class discrepancy of different categories (GAN-generated or pristine images) and improve intra-class compactness from different domains (source attributions) in the same category. Extensive experiments are conducted which consider various cross-domain scenarios to verify the generalization of the proposed method. Besides, our method exhibits satisfying robustness against common post-processings, even when data augmentation is not considered during the training stage.

MKN: Metakernel Networks for Few Shot Remote Sensing Scene Classification

Few-shot remote sensing scene classification tries to make a model quickly adapt to new scenes with only a few samples that do not appear in the closed training set. Since limited samples can hardly describe the distribution of data, it is a challenge for a model to learn good generalized features. Since limited samples are rarely representative, it is another challenge for a model to learn classification boundaries that depend on sample bias. Therefore, we propose a method called metakernel networks (MKNs) to solve the challenges via integrating a parametric linear classifier (PLC) into the metalearning framework to address the former problem assembling a metakernel strategy (MKS) and a stretching loss (Stloss) to address the later problem. The PLC learns prior knowledge from different tasks sampled from the same task family to learn rich features. The MKS is designed to remap low-dimensional indistinguishable features to a high-dimensional space to solve the low-dimensional feature entanglement caused by large intraclass differences between remote sensing images. The Stloss reduces the dependence of the hyperplane formed by a few points in each class on sample selection by reducing the intraclass and interclass variance ratios, thus solving the boundary fragility problem caused by interclass similarity. Experiments on three public datasets, UC_Merced, NWPU-RESISC45, and AID, show the state-of-the-art performance by the accuracy improvement of 5.04%, 4.81%, and 4.69% respectively. Our findings provide a new perspective by suggesting that the previously neglected issue of classification boundaries may be a key factor for few-shot remote sensing scene classification.

Resource allocation for joint energy and spectral efficiency in cloud radio access network based on deep reinforcement learning

AbstractThe rapid increase of user data traffic demand has promoted the telecommunication sector toward adopting a new generation, that is, fifth‐generation (5G). Cloud radio access network (CRAN) has gained considerable attention to satisfy the high traffic demand in the 5G network via deployment and intelligent management of multiple remote radio units (RRHs). However, optimizing the instantaneous network performance may lead to myopic decision‐making, such as excessive on/off switching of RRHs. This paper proposes a deep reinforcement learning (DRL) based framework, with the goal of maximizing the long‐term tradeoff between energy efficiency (EE) and spectral efficiency (SE). To this end, we specifically formulate the joint optimization problem as a Markov decision process (MDP) subject to the constraints on per‐RRH transmission power and user quality of service (QoS) demands. Meanwhile, considering the spatio‐temporal channel state information (CSI), we adopt machine learning (ML) techniques to extract generalized features before feeding them into the input of DRL. Combining with RRH status and QoS requirements, the proposed algorithm can learn the near‐optimal control strategy to turn on/off RRHs, followed by solving a power optimization problem. Simulation results reveal that the proposed method yields better performance as compared to myopic and DRL methods without considering CSI generalization.

A novel weather prediction model using a hybrid mechanism based on MLP and VAE with fire-fly optimization algorithm

The future weather data source will continue to grow rapidly, and new developments in machine learning would allow government agencies and companies to use all this data further. The weather prediction will never be flawless, but artificial intelligence (AI) can strive to enhance the exactness and consistency of the process. This paper proposes an approach using a hybrid mechanism based on multi-layer perceptron (MLP) and variational auto-encoder (VAE) with a fire-fly optimization mechanism. Weather-related data contains many features. A few of which are global or generalized features, and some are local or internal features. Single mechanism may not be effective in the process of extracting the specified features. Hence, a hybrid mechanism with the support of VAE and MLP is proposed to extract features and do classification. VAE is used to extract the global features from the weather data and the obtained or processed intermediately output given to the input as the MLP, which will extract all local or internal features very effectively.

Open-Set Domain Adaptation in Machinery Fault Diagnostics Using Instance-Level Weighted Adversarial Learning

Data-driven machinery fault diagnosis methods have been successfully developed in the past decades. However, the cross-domain diagnostic problems have not been well addressed, where the training and testing data are collected under different operating conditions. Recently, domain adaptation approaches have been popularly used to bridge this gap, which extract domain-invariant features for diagnostics. Despite the effectiveness, most existing methods assume the label spaces of training and testing data are identical that indicates the fault mode sets are the same in different scenarios. In practice, new fault modes usually occur in testing, which makes the conventional methods focusing on marginal distribution alignment less effective. In order to address this problem, a deep learning-based open-set domain adaptation method is proposed in this study. Adversarial learning is introduced to extract generalized features, and an instance-level weighted mechanism is proposed to reflect the similarities of testing samples with known health states. The unknown fault mode can be effectively identified, and the known states can be also recognized. Entropy minimization scheme is further adopted to improve generalization. Experiments on two practical rotating machinery datasets validate the proposed method. The results suggest the proposed method is promising for open-set domain adaptation problems, which largely enhances the applicability of data-driven approaches in the real industries.

SAR Image Change Detection Based on Semisupervised Learning and Two-Step Training

Change detection, as an essential part of synthetic aperture radar (SAR) automatic target recognition (ATR) systems, remains a challenging problem. The lack of labeled data seriously impedes the development of deep learning-based methods in SAR applications. In this letter, we propose a patch-based semisupervised method to detect changed pixels from limited training data in the field of SAR. The complete approach includes the unsupervised pretraining stage and the iterative discrimination stage. The states of pixels are determined by comparing the corresponding image patches. First, to reduce the impact of insufficient data, we train a trapezium U-Net-like structure to extract representative and generalized features by unsupervised training on all pixels. Feature maps contain detailed and semantic information due to repetitive feature fusions. A designed feature activation module is utilized to recalibrate fused features. Finally, the discrimination is finished by a simple classification convolutional neural network (CNN) following a two-step training strategy. Ablation studies indicate the function provided by the proposed modifications. Experiments on real SAR images demonstrate that the proposed method can improve the detection accuracy by more than 1.2% compared with other state-of-the-art deep learning-based methods.

DeepQoE: A Multimodal Learning Framework for Video Quality of Experience (QoE) Prediction

Recently, many models have been developed to predict video Quality of Experience (QoE), yet the applicability of these models still faces significant challenges. Firstly, many models rely on features that are unique to a specific dataset and thus lack the capability to generalize. Due to the intricate interactions among these features, a unified representation that is independent of datasets with different modalities is needed. Secondly, existing models often lack the configurability to perform both classification and regression tasks. Thirdly, the sample size of the available datasets to develop these models is often very small, and the impact of limited data on the performance of QoE models has not been adequately addressed. To address these issues, in this work we develop a novel and end-to-end framework termed as DeepQoE. The proposed framework first uses a combination of deep learning techniques, such as word embedding and 3D convolutional neural network (C3D), to extract generalized features. Next, these features are combined and fed into a neural network for representation learning. A learned representation will then serve as input for classification or regression tasks. We evaluate the performance of DeepQoE with three datasets. The results show that for small datasets (e.g., WHU-MVQoE2016 and Live-Netflix Video Database), the performance of state-of-the-art machine learning algorithms is greatly improved by using the QoE representation from DeepQoE (e.g., 35.71% to 44.82%); while for the large dataset (e.g., VideoSet), our DeepQoE framework achieves significant performance improvement in comparison to the best baseline method (90.94% vs. 82.84%). In addition to the much improved performance, DeepQoE has the flexibility to fit different datasets, to learn QoE representation, and to perform both classification and regression problems. We also develop a DeepQoE based adaptive bitrate streaming (ABR) system to verify that our framework can be easily applied to multimedia communication service. The software package of the DeepQoE framework has been released to facilitate the current research on QoE.

Insights into the evolution of the thermal field in evaporating sessile pure water drops

We investigate the evolution of the thermal field during evaporation, a fundamental aspect of evaporating sessile drops. With numerous reports in the literature investigating the contact line dynamics, we aspire to identify generalized features in the evolution of the thermal field and ultimately correlate these with the contact line dynamics. Considering a broad range of experimental parameters such as substrate wettability, substrate temperature, initial volume of the drop, and ambient relative humidity results in a wide range of evaporation rates, in turn affecting the strength of internal convective flows. Infrared thermography is utilized to extract the thermal field at the liquid–vapor interface, and optical imaging is used to record the evolution of drop shape during evaporation. We observe that the onset and presence of a convective cell as a cold spot at the interface highlights a non-axisymmetry in the thermal field. In consequence, a hitherto unreported asymmetry in the internal flow field is observed, as evidenced by the particle image velocimetry. Among the multitude of experiments conducted, we report four distinct trends in the evolution of interfacial temperature difference depending on the presence and duration of the presence of the convective cell, which are elucidated by discussing the evolution of maximum and minimum temperatures at the interface. The interplay between heat conducted into the drop and heat released due to evaporation can result in a momentary decrease in temperature of the drop, which is not reported previously. Lastly, a theoretical estimate for the temperature difference within the drop is extracted using vapor diffusion model and energy balance during evaporation. Comparison of this theoretical temperature difference with experimental observations highlights the influence of internal convective flows in homogenizing the thermal field within the drop.

Robust dialog state tracker with contextual-feature augmentation

Dialog state tracking (DST), which estimates dialog states given a dialog context, is a core component in task-oriented dialog systems. Existing data-driven methods usually extract features automatically through deep learning. However, most of these models have limitations. First, compared with hand-crafted delexicalization features, such features in deep learning approaches are not universal. However, they are important for tracking unseen slot values. Second, such models do not work well in situations where noisy labels are ubiquitous in datasets. To address these challenges, we propose a robust dialog state tracker with contextual-feature augmentation. Contextual-feature augmentation is used to extract generalized features; hence, it is capable of solving the unseen slot value tracking problem. We apply a simple but effective deep learning paradigm to train our DST model with noisy labels. The experimental results show that our model achieves state-of-the-art scores in terms of joint accuracy on the MultiWOZ 2.0 dataset. In addition, we show its performance in tracking unseen slot values by simulating unseen domain dialog state tracking.

DenseHyper: an automatic recognition system for detection of hypertensive retinopathy using dense features transform and deep-residual learning

High blood pressure and diabetes are associated with a retinal abnormality known as Hypertensive Retinopathy (HR). The severity-level and duration of hypertension are straightly related to the incidence of HR-eye disease. The HR damages the pathological lesions of eyes such as arteriolar narrowing, retinal hemorrhage, macular edema, cotton wool spots, and blood vessels. In the early stages, it is important to detect and diagnose HR to prevent eye blindness. Currently, there are few computerize systems developed to recognize HR. However, those systems focused on extracting features through hand-craft and deep-learning models (DLMs) based techniques. As a result, the complex image processing algorithms are required in case of hand-crafted features and it is difficult to define generalized features by DLMs to recognize HR. Moreover, the classification accuracy is not up-to-the-mark even though by using deep-feature techniques as observed in state-of-the-art HR diagnostics systems. To solve these problems, a novel hypertensive retinopathy (DenseHyper) system is developed to detect the HR based on a proposed trained features layer (TF-L) and dense feature transform layer (DFT-L) to the deep residual learning (DRL) methods. The DenseHyper system consists of different multilayer dense architecture by integrating of TF-L by convolutional neural network (CNN) to learn features from different lesions, and generate specialized features by DFT-L. To develop DenseHyper system, a learning based dense feature transform (DFT) approach was integrated to increase classification accuracy. Three online sources besides one private data are gathered to test and compare the DenseHyper system. To show the performance of the DenseHyper system, the statistical analysis is also performed on 4270 retinal fundus images through sensitivity (SE), specificity (SP), accuracy (ACC) and area under the receiver operating curve (AUC) metrics. The significant results were achieved compare to state-of-the-art methods. On average, the SE of 93%, SP of 95%, ACC of 95% and 0.96 of AUC values were obtained through a 10-fold cross-validation test. Experimental results confirm the applicability of the DenseHyper system to accurately diagnosis of hypertensive retinopathy.

Speech emotion recognition using fusion of three multi-task learning-based classifiers: HSF-DNN, MS-CNN and LLD-RNN

Speech emotion recognition plays an increasingly important role in emotional computing and is still a challenging task due to its complexity. In this study, we developed a framework integrating three distinctive classifiers: a deep neural network (DNN), a convolution neural network (CNN), and a recurrent neural network (RNN). The framework was used for categorical recognition of four discrete emotions (i.e., angry, happy, neutral and sad). Frame-level low-level descriptors (LLDs), segment-level mel-spectrograms (MS), and utterance-level outputs of high-level statistical functions (HSFs) on LLDs were passed to RNN, CNN, and DNN, separately. Three individual models of LLD-RNN, MS-CNN, and HSF-DNN were obtained. In the models of MS-CNN and LLD-RNN, the attention mechanism based weighted-pooling method was utilized to aggregate the CNN and RNN outputs. To effectively utilize the interdependencies between the two approaches of emotion description (discrete emotion categories and continuous emotion attributes), a multi-task learning strategy was implemented in these three models to acquire generalized features by simultaneously operating classification of discrete categories and regression of continuous attributes. Finally, a confidence-based fusion strategy was developed to integrate the power of different classifiers in recognizing different emotional states. Three experiments on emotion recognition based on the IEMOCAP corpus were conducted. Our experimental results show that the weighted pooling method based on attention mechanism endowed the neural networks with the capability to focus on emotionally salient parts. The generalized features learned in the multi-task learning helped the neural networks to achieve higher accuracies in the tasks of emotion classification. Furthermore, our proposed fusion system achieved weighted accuracy of 57.1% and unweighted accuracy of 58.3%, which were significantly higher than those of each individual classifier. The effectiveness of the proposed approach based on classifier fusion was thus validated.

Deep Neural Network with Joint Distribution Matching for Cross-Subject Motor Imagery Brain-Computer Interfaces

Motor imagery brain-computer interfaces (BCIs) have demonstrated great potential and attract world-spread attentions. Due to the nonstationary character of the motor imagery signals, costly and boring calibration sessions must be proceeded before use. This prevents them from going into our realistic life. In this paper, the source subject's data are explored to perform calibration for target subjects. Model trained on source subjects is transferred to work for target subjects, in which the critical problem to handle is the distribution shift. It is found that the performance of classification would be bad when only the marginal distributions of source and target are made closer, since the discriminative directions of the source and target domains may still be much different. In order to solve the problem, our idea comes that joint distribution adaptation is indispensable. It makes the classifier trained in the source domain perform well in the target domain. Specifically, a measure for joint distribution discrepancy (JDD) between the source and target is proposed. Experiments demonstrate that it can align source and target data according to the class they belong to. It has a direct relationship with classification accuracy and works well for transferring. Secondly, a deep neural network with joint distribution matching for zero-training motor imagery BCI is proposed. It explores both marginal and joint distribution adaptation to alleviate distribution discrepancy across subjects and obtain effective and generalized features in an aligned common space. Visualizations of intermediate layers illustrate how and why the network works well. Experiments on the two datasets prove the effectiveness and strength compared to outstanding counterparts.

A Dual-Channel convolution neural network for image smoke detection

Image smoke detection is a challenging task due to the difference of color, texture, and shape of smoke. In recent years, deep learning has greatly improved the performance of image classification and detection. In this paper, we propose a Dual-Channel Convolutional Neural Network (DC-CNN) using transfer learning for detecting smoke images. Specifically, an AlexNet network with transfer learning, used to extract generalized features, is designed on the first channel as the main framework of entire network. The second channel is a tidy convolution neural network for extracting specific and detailed features. To guarantee the robustness of the network, two channels of the network are trained separately and their features are fused in the concat layer. The experimental data sets consist of smoke images and non-smoke images, and some challenging non-smoke images are added into the data sets as a supplement. Experimental results show that the proposed method can work effectively and achieve detection rate above 99.33%.

Interrelation Of Labor Motivation and Adaptation Potential of Interior Ministery Members

The article is devoted to the problem of labour motivation and adaptation potential of law enforcement officials. It examines the concepts of «motivation», «labour motivation», «adaptation» and «adaptation potential». The article presents a study which purpose is to investigate the interrelationship of labor motivation and the adaptive potential of Interior Ministry members. To achieve this goal, the article uses methods designed to assess the personal adaptive abilities, taking into account socio-psychological and some psychophysiological characteristics that reflect generalized features of neuropsychic and social development, to measure the subjective feeling of personal effectiveness, to identify the motivation for success and the motivation for fear of failure, and also a questionnaire to identify types of labor motivations. The article carries out an empirical study of the interrelationship of labor motivation and the adaptive potential of members of the Ministry of Internal Affairs, as well as reveals positive correlation dependencies between the labor motivation and the adaptation potential of employees of the Internal Ministry members.

WS-AM: Weakly Supervised Attention Map for Scene Recognition

Recently, convolutional neural networks (CNNs) have achieved great success in scene recognition. Compared with traditional hand-crafted features, CNN can be used to extract more robust and generalized features for scene recognition. However, the existing scene recognition methods based on CNN do not sufficiently take into account the relationship between image regions and categories when choosing local regions, which results in many redundant local regions and degrades recognition accuracy. In this paper, we propose an effective method for exploring discriminative regions of the scene image. Our method utilizes the gradient-weighted class activation mapping (Grad-CAM) technique and weakly supervised information to generate the attention map (AM) of scene images, dubbed WS-AM—weakly supervised attention map. The regions, where the local mean and the local center value are both large in the AM, correspond to the discriminative regions helpful for scene recognition. We sampled discriminative regions on multiple scales and extracted the features of large-scale and small-scale regions with two different pre-trained CNNs, respectively. The features from two different scales were aggregated by the improved vector of locally aggregated descriptor (VLAD) coding and max pooling, respectively. Finally, the pre-trained CNN was used to extract the global feature of the image in the fully- connected (fc) layer, and the local features were combined with the global feature to obtain the image representation. We validated the effectiveness of our method on three benchmark datasets: MIT Indoor 67, Scene 15, and UIUC Sports, and obtained 85.67%, 94.80%, and 95.12% accuracy, respectively. Compared with some state-of-the-art methods, the WS-AM method requires fewer local regions, so it has a better real-time performance.