Search

• Pinch force strength is accurately predicted using sEMG. • Different pinch movements are accurately classified using sEMG. • Similarity of EMG features is evaluated. • An EMG feature-to-image method is proposed. The movement of intelligent EMG-driven prosthesis mainly relies on the synergy of different fingers to achieve function of grasping objects. The paper proposes a novel scheme for force prediction and movement classification about pinch between different fingers based on surface electromyography (sEMG) using machine learning. The pinch force and sEMG signals are recorded synchronously by a data acquisition device. Eight features are extracted, which are proven to have better performance in the estimation of sEMG-to-force. We present a novel feature selection method that uses the one-dimensional time series similarity assessment based on Manhattan distance to eliminate the repetitive information between features. Three optimal features carrying less repetitive information are retained. Seven machine learning algorithms are used to predict force strength. The results show that the Long Short Term Memory (LSTM) has the best performance of force prediction, achieving a R 2 of 0.9517 and RMSE of 3.2723. The paper proposes a novel method of converting EMG feature sequences to the normalized gray image in order to classify the finger movement. Five classifiers based on image feature extraction and the Convolutional Neural Network (CNN) are developed respectively. The experimental results indicate that the CNN performs best, achieving an accuracy of 97.66%. In this way, it not only realizes the accurate force prediction, but also realizes the movement classification between different fingers. The proposed methodology has the potential to realize simultaneous force and movement control of prosthetic hand. [ABSTRACT FROM AUTHOR]

The aging of transformer oil–paper insulation (OPI) contributes significantly to transformer failure and reduced service life. Therefore, reliable and precise evaluation of the aging status of transformer insulation condition is crucial in averting transformer failure. This article proposes a novel deep learning-based aging condition assessment technique of transformers by combining FDS (Frequency domain spectroscopy) and squeeze-and-excitation-enabled convolutional neural network (SE-CNN). An intrinsic product-based transformation using a Gramian angular field is employed to dimensionally expand the 1-D FDS signals acquired from OPI samples into 2D images. Each feature extraction channel in the SE-CNN-based aging diagnosis framework is assigned a specific weight, enabling the model to concentrate on the most relevant features while discarding the rest. Furthermore, the effect of moisture variation on the model is also considered under three different conditions to ensure the consequence of compensation for the effect of moisture on FDS. A comparison with traditional machine learning models has also been made to ensure the viability of the proposed methodology. Additionally, the model is interpreted by creating explanation weight plots of the feature coefficients using a local interpretable model-agnostic explanations (LIME) method in Python. The findings demonstrate that the novel aging diagnosis framework visually assesses the transformer's OPI system with the highest accuracy of 99.12% in the test set. Furthermore, it facilitates the automatic extraction of deep aging features, thus providing better generalization and convergence capability. [ABSTRACT FROM AUTHOR]

Longitudinal impulse, as a common dynamic problem of heavy haul trains, could induce the lateral instability of the coupler and the separation of the locomotive and wagon connection and thus seriously hinder the train running safety. The coupler force undergoes complex changes during longitudinal impulse transmission and exhibits strong nonlinear characteristics, which poses significant challenges to the identification of the coupler force. To address these issues, this paper reports a novel quantitative method for the coupler force identification of heavy haul locomotives. Firstly, the moving average filtering method is used to preprocess the signal of longitudinal relative displacement between the coupler and carbody (LRDCC), and the approximate relative velocity (ARV) feature is also extracted from the preprocessed signal. Then, a Convolutional Neural Network (CNN) model is constructed to extract high-level features from LRDCC and ARV. The model with randomly initialized weight parameters does not require additional training and thus reduces the computational cost. Finally, an Extreme Learning Machine (ELM) is used as the regression layer to process the extracted high-level features and further improve the identification performance. Additionally, the dung beetle algorithm is used to obtain the optimal model parameters to improve the coupler force identification performance. Not only a dynamics simulation but a field test is carried out to show that the proposed method could accurately detect the coupler force of heavy haul locomotives. Furthermore, the proposed method shows promising prospects in replacing the traditional monitoring method under the actual long-term working conditions of trains. • An indirect measurement scheme is proposed for coupler force dynamic identification. • The proposed network model presents high identification Accuracy and efficiency. • Method showcases strong generalization under complex, long-term train operations. [ABSTRACT FROM AUTHOR]

Implementing key engineering solutions to optimise the operation of energy industries requires daily electricity demand forecasting and including uncertainty, to promote markets insight analysis as part of their strategic planning, regulating and supplying electricity to consumers. This paper proposes hybrid artificial intelligence models combining convolutional neural networks (CNN) as a feature extraction algorithm with extreme learning machines (ELM) as a framework to predict electricity demand with confidence intervals generated by Kernel Density Estimation (KDE) approaches. In order to develop CELM-KDE model, time-lagged series of daily electricity demand with local climate variables based on the air temperature, atmospheric vapour pressure, evaporation, solar radiation, humidity and sea level pressures are used to train the proposed CELM-KDE hybrid model. In order to fully evaluate the newly developed model from a point-based, as well as a probabilistic prediction strategy, the observed and predicted electricity demand as well as the probability distribution of errors are analysed using KDE method that operates without prior data distribution assumptions. Based on observed and predicted electricity demand and the relevant probabilistic confidence intervals generated by the CELM-KDE model, the final results show that the proposed method attains significantly better probability interval predictions than traditionally-used point-based models. The proposed CELM-KDE model is demonstrated to be highly effective in providing a comprehensive coverage of predicted errors, as well as providing greater insights into the average bandwidth and detailed predicted electricity demand in the testing stage. The results also indicate that the proposed hybrid model is a reliable decision support tool to develop engineering solutions in area of energy modelling, monitoring and forecasting, which could potentially be useful to the industry policymakers. We show that the point-and probabilistic-based electricity demand predictive models can be employed as an effective tool to improve accuracy of forecasting and provision of insights for national electricity markets and key energy industry stakeholder application tools. [Display omitted] • A hybrid deep learning model for daily electricity demand prediction is proposed. • A Kernel Density Estimation is applied to derive probabilistic prediction intervals. • The algorithm quantifies uncertainties in point-based electricity demand predictions. • The proposed model outperforms deep learning models using statistical uncertainty indicators. [ABSTRACT FROM AUTHOR]

[Display omitted] • A corn quality detection method based on spectral technology and deep learning was proposed. • This method solves the time-consuming and labor-intensive problem of corn quality detection. • This paper provides a new idea and method for corn quality detection. • The research results provide technical support for the rapid detection of crop quality. Corn is an important food crop in the world. With economic development and population growth, the nutritional quality of corn is of great significance to high-quality breeding, scientific cultivation and fine management. Aiming at the problems of cumbersome steps, time-consuming and laborious, and low accuracy in the current research on corn quality detection. This paper proposes to combine near-infrared (NIR) spectroscopy technology with deep learning technology to build a corn quality detection model based on convolutional neural network (LeNet-5). The original spectral data were preprocessed by wavelet transform (WT) and multivariate scattering correction (MSC) to remove noise interference and spectral scattering information. The Competitive Adaptive Reweighted Sampling Algorithm (CARS) was applied to optimize the characteristic wavenumber and reduce redundant data. According to the optimized characteristic wave number, it was input into the constructed corn quality detection model for simulation test, and the average detection accuracy rate of the test set was 96.46%, the average precision rate was 95.42%, the average recall rate was 97.92%, the average F1score was 96.64%, and the average recognition time was 51.95 s. Compared with traditional machine learning models such as BP neural network, K Nearest Neighbor (KNN), Support Vector Machine (SVM), Generalized Linear Model (GLM), Linear Discriminant Analysis (LDA), and Naive Bayesian (NB), the deep learning LeNet-5 network model constructed in this paper has an average accuracy increase of 39.32%, and has a higher detection accuracy. [ABSTRACT FROM AUTHOR]

Automatic fault diagnosis in many systems is performed via the analysis of vibration data in the time or frequency domain. In the literature, many approaches for extracting featured from signals for fault diagnosis have been proposed. In this paper, we apply a variety of transform functions including Fourier, Wavelet, etc. to extract features from the vibration data, which includes the statistical features and some features automatically extracted via Convolutional Neural Networks (CNNs). For each of these feature extraction approaches, a learning algorithm is trained to diagnose the faults and their results are aggregated in an ensemble machine learning algorithm. The weights of the base learner algorithms are optimized via an evolutionary algorithm to achieve the best-weighted voting scheme. The architecture of CNNs has a significant effect on the performance of the algorithm, thus, in this paper, an evolutionary algorithm is proposed to find the best architecture for CNNs in fault diagnosis. The CNNs are trained via gradient descent algorithms which suffer from getting stuck in local optima. To manage this, we propose an evolutionary algorithm that benefits from the speed of gradient descent and the global search of evolutionary algorithms. The proposed algorithm is tested on a number of benchmark problems and the experimental results are presented. • An ensemble algorithm is proposed for fault diagnosis. • Different feature extractors are exploited. • An evolutionary training algorithm is proposed. [ABSTRACT FROM AUTHOR]

Convolutional Neural Networks have dominated the field of computer vision for the last ten years, exhibiting extremely powerful feature extraction capabilities and outstanding classification performance. The main strategy to prolong this trend in the state-of-the-art literature relies on further upscaling networks in size. However, costs increase rapidly while performance improvements may be marginal. Our main hypothesis is that adding additional sources of information can help to increase performance and that this approach is more cost-effective than building bigger networks, which involve higher training time, larger parametrisation space and higher computational resources requirements. In this paper, an ensemble method for accurate image classification is proposed, fusing automatically detected features through a Convolutional Neural Network and a set of manually defined statistical indicators. Through a combination of the predictions of a CNN and a secondary classifier trained on statistical features, a better classification performance can be achieved cheaply. We test five different CNN architectures and multiple learning algorithms in a diverse number of datasets to validate our proposal. According to the results, the inclusion of additional indicators and an ensemble classification approach help to increase the performance in all datasets. Both code and datasets are publicly available via GitHub at: https://github.com/jahuerta92/cnn-prob-ensemble. • A modular architecture to improve out-of-the-box computer vision CNNs. • A fusion of CNNs and image features that boosts classification performance and explainability. • Thorough evaluation over several datasets, state-of-the-art CNNs, machine learning algorithms and their combinations. • Results show that adding this fusion approach improves over the baselines frequently. [ABSTRACT FROM AUTHOR]

The presence of Beggiatoa Spp. indicates anoxic conditions or 'poor condition' in marine sediments beneath aquaculture pens, resulting from organic enrichment. Currently, the most efficient approach to estimate Beggiatoa Spp. coverage, and thus the extent of the issue, involves video surveys which are scored by human observers for presence of this bacteria. However, this approach is highly time-consuming and relies heavily on the expertise and experience of the individuals involved, thus affecting its accuracy. Machine learning-based computer vision techniques, such as Convolutional Neural Networks (CNNs), offer the potential for automated estimation of Beggiatoa Spp. coverage. However, most existing machine learning methods focus solely on the estimation of the coverage via presence, absence of a single type of Beggiatoa Spp.. These approaches typically rely on binary classification to distinguish the object from the background when estimating coverage. Nevertheless, the inclusion of subordinate categories within high-level classifications poses a great challenge for accurately estimating their coverage rates. In this paper, an adaptive ensemble learning approach was proposed to estimate Beggiatoa Spp. coverage. Unlike other approaches, the proposed approach is capable in adaptively extracting and fusing features from underwater images and accurately estimating the coverages of multiple types of Beggiatoa Spp. through ensemble learning. Experimental results demonstrated that our proposed approach outperforms other approaches in terms of both effectiveness and efficiency in estimating Beggiatoa Spp. coverage. • Adaptive feature extraction and fusion enhance relevant and informative features' identification. • Adaptive feature extraction and fusion enhance feature representation and reduce time cost. • Flexible ensemble learning framework improves accuracy, robustness, and efficiency. • Flexible ensemble learning framework can be extended to other models or applications. • Novel machine learning-based approach estimates subordinate categories of Beggiatoa Spp. coverage. [ABSTRACT FROM AUTHOR]

A novel convolutional neural network is proposed for local prior feature embedding and imbalanced dataset modeling for multi-channel time-varying signal classification. This model consists of a single-channel signal feature parallel extraction unit, a multi-channel signal feature integration unit, a local feature embedding and feature similarity measurement unit, a full connection layer, and a Softmax classifier. An algorithm combining dynamic clustering and sliding window was used to select segments signals with typical local features in each pattern class, forming a typical local feature set. The one-dimensional CNNs were used to extract features from the single-channel signal in parallel, a comprehensive feature matrix of the multi-channel signal and the local feature matrix templates were produced. Using the method of external embedding, based on the sliding window and dynamic time warping (DTW) algorithm, the local feature similarities between the local feature template of each pattern class and the comprehensive feature sub-matrix of the input signal were measured, and the maximum values were selected to construct a local feature similarity vector in order. The information fusion was realized through a full connection layer. The proposed methodology can extract and represent both global and local signals features, strengthen the role of prior local feature in classification and improve the modeling properties of imbalanced datasets. A comprehensive learning algorithm is presented in this paper. The classification diagnosis of cardiovascular disease based on 12-lead ECG signals was used as a verification experiment. Results showed that the accuracy and generalization for the proposed technique were significantly improved. • Local prior feature knowledge embedding, model structure and data constraints. • The proposed model has the properties of feature combination and fusion analysis. • Improve the distinguishing ability of different pattern features. • Model small-scale and imbalanced datasets and maintain diversity of signal features. [ABSTRACT FROM AUTHOR]

The involvement of external vendors in semiconductor industries increases the chance of hardware Trojan (HT) insertion in different phases of the integrated circuit (IC) design. Recently, several partial reverse engineering (RE) based HT detection techniques are reported, which attempt to reduce the time and complexity involved in the full RE process by applying machine learning or image processing techniques in IC images. However, these techniques fail to extract the relevant image features, not robust to image variations, complicated, less generalizable, and possess a low detection rate. Therefore, to overcome the above limitations, this paper proposes a new partial RE based HT detection technique that detects Trojans from IC layout images using Deep Convolutional Neural Network (DCNN). The proposed DCNN model consists of stacking several convolutional and pooling layers. It layer-wise extracts and selects the most relevant and robust features automatically from the IC images and eliminates the need to apply the feature extraction algorithm separately. To prevent the over-training of the DCNN model, a new stopping condition method and two new metrics, namely Accuracy difference measure (ADM) and Loss difference measure (LDM), are proposed that halts the training only when the performance of our model genuinely drops. Further, to combat the issue of process variations and fabrication noise generated during the RE process, we include noisy images with varying parameters in the training process of the model. We also apply the data augmentation and regularization techniques in the model to address the issues of underfitting and overfitting. Experimental evaluation shows that the proposed technique provides 99% and 97.4% accuracy on Trust-Hub and synthetic ISCAS dataset, respectively, which is on-an-average 15.83% and 21.69% higher than the existing partial RE based techniques. • A new partial RE based HT detection technique using the DCNN model has been proposed. • A new DCNN model architecture is proposed to perform automatic and relevant feature extraction from IC images. • Proposed two new metrics and stopping condition method avoid overtraining of the DCNN model effectively. • The proposed technique provides high detection accuracy on Trust-Hub and synthetic ISCAS benchmarks. [ABSTRACT FROM AUTHOR]

The Internet of Things (IoT) has experienced phenomenal expansion over the past few years and has emerged as one of the most dynamic sectors of the international market. Android has become the driving force behind the rapid development of the IoT, and malware is one of the most severe concerns. Android is an open-source platform with a significant number of users. The identification of Android malware has become more critical in cybersecurity due to the meteoric rise of Android malware over the past few years and the severity of the damage it inflicts on smartphone users. Existing traditional Android malware detection methods based on machine learning have achieved encouraging performance. Still, the solution requires a massive workforce of feature engineering to build dynamic or static features and continuous learning through pre-extracted features to maintain high performance in identifying malware. Therefore, technologies for end-to-end malware detection without human interaction are necessary. In response to the rapid evolution of Android malware, a convolutional neural network that can acquire knowledge without requiring feature extraction is ideally suited. This paper offers a unique classification approach for malware based on a convolutional neural network, with feature extraction selected from the current pre-trained models DenseNet169, Xception, InceptionV3, ResNet50, and VGG16. The proposed methods have an advantage over existing detection methods due to their end-to-end learning process. Our suggested approaches convert Android APK files to binary codes and RGB images for usage as inputs to deep learning models. These models are trained and evaluated using the CICInvesAndMal2019 dataset, which consists of 333 benign and 428 malicious applications. Experiments demonstrate that the proposed approaches achieve a classification accuracy of 95.24%, 95.24%, and 95.83% for DenseNet169, InceptionV3, and VGG16, respectively, on a 761-sample dataset of malware, which is superior to other compared algorithms. Compared to previous methods, our suggested methods are more applicable to Android IoT devices since they do not require manual feature engineering and consume fewer resources. [ABSTRACT FROM AUTHOR]

The assessment of peak deviatoric stress (q peak) in gravels is crucial for engineering stability and safety. Due to the various influencing factors and the limited available samples, present empirical methods and conventional machine learning algorithms have limitations in generalization capability and achieving high performance in this complex relationship. This paper proposes a novel deep learning-based model termed Seft-Net. The model utilizes a segmented feature transformation (Seft) approach to improve convergence and robustness by leveraging prior knowledge that grouping gravel properties into three categories: particle (e.g., shape, hardness), soil mass (e.g., gradation, void ratio), and external factors (e.g., confining pressure). Subsequently, it applies a feature transformation module using depthwise separable convolutions to learn and fuse hierarchical representations across multiple layers. This enables Seft-Net to automatically extract high-level features from gravel properties that capture intricate nonlinear relationships with q peak. The model is trained and validated using a large-scale practical dataset from triaxial test results. Ablation studies confirmed the superiority of the proposed segmented feature transformation approach over the unsegmented approach, and the feature transformation module demonstrated optimal feature extraction capabilities. Comparative evaluations with three machine learning methods further validated the superior performance of Seft-Net. [ABSTRACT FROM AUTHOR]

In semiconductor manufacturing, wafer map defect pattern provides critical information for facility maintenance and yield management, so the classification of defect patterns is one of the most important tasks in the manufacturing process. In this paper, we propose a novel way to represent the shape of the defect pattern as a finite-dimensional vector, which will be used as an input for a neural network algorithm for classification. The main idea is to extract the topological features of each pattern by using the theory of persistent homology from topological data analysis (TDA). Through some experiments with a simulated dataset, we show that the proposed method is faster and much more efficient in training with higher accuracy, compared with the method using convolutional neural networks (CNN) which is the most common approach for wafer map defect pattern classification. Moreover, it was shown that our method outperforms the CNN-based method when the number of training data is not enough and is imbalanced. • A novel approach for wafer map defect pattern classification is proposed. • This new method is based on the theory from topological data analysis. • The proposed method is accurate and computationally efficient. • This algorithm is fast enough to provide a real-ti me prediction. • This method is robust against a small and imbalanced dataset. [ABSTRACT FROM AUTHOR]

• An automated computer aided diagnosis system using deep learning is proposed. • The proposed system classifies thyroid histopathology images into two classes. • Transfer learning is applied to 5 different pre-trained convolutional neural networks. • Data augmentation is employed to improve the diversity of dataset. • Results show the success of transfer learning to classify thyroid histopathology images. CAD systems for histopathology image analysis using machine learning is a well researched subject. Deep learning is playing a major role in advancing this research in the recent years. This paper presents an automated thyroid histopathology image classification system with deep neural networks using the theory of transfer learning and popular pre-trained convolutional neural networks (CNNs). In this experiment-based study, two forms of transfer learning namely feature extraction and fine tuning are applied on popular state-of-the-art CNN architectures such as VGGNet, ResNet, InceptionNet and DenseNet to classify thyroid histopathology images. Accuracy, precision, sensitivity, specificity, area under receiver operating characteristic (AUROC) analysis and F1-score are used to evaluate the performance of the architectures. The results are promising and demonstrate the feasibility of transfer learning for thyroid histopathology image analysis. [ABSTRACT FROM AUTHOR]

This paper aims to address the human action recognition issue by using convolutional long short-term memory networks (Conv-LSTM) and fully-connected LSTM (FC-LSTM) with different attentions. To this end, the spatial-temporal dual-attention network (STDAN), which is mainly composed of feature extraction, attention and fusion modules, is designed. Different from the features of high-level fully-connected layer mostly used in previous work, the features of convolution and fully-connected layers of convolutional neural network (CNN) are both extracted in STDAN, which can enrich the initial level of video representation. Besides, the Conv-LSTM and FC-LSTM are employed to handle the long-duration sequential features with different temporal context information. To reinforce the spatial-temporal attention ability, a temporal attention module (TAM) and a joint spatial-temporal attention module (JSTAM) are implemented. Through the principle components analysis (PCA) and features fusion, the potential of STDAN is effectively explored and weighted. Finally, the experimental results show that the proposed STDAN has better recognition performance than existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Convolutional Neural Networks (CNN) offer promising opportunities to automatically glean scientifically relevant information directly from annotated images, without needing to handcraft features for detection. Crater counting started with hand counting hundreds, thousands, or even millions of craters in order to determine the age of geological units on planetary bodies of the solar system. Automated crater detection algorithms have attempted to speed up this process. Previous research has employed computer vision techniques with handcrafted features such as light and shadow patterns, circle finding, or edge detection. This research continues, but now some researchers use techniques like convolutional neural networks that enable the algorithm to develop its own features. As the field of machine learning undergoes exponential growth in terms of paper count and research methods, the crater counting application can benefit from the new research, especially when conducting joint interdisciplinary projects. Despite these advancements, the crater counting community has not yet adopted standard methods for automating the process despite decades of research. This survey enumerates challenges for both planetary geologists and machine learning researchers, looks at the recent automatic crater detection advancements using machine learning techniques (primarily in methods using CNNs), and makes recommendations for the path toward greater automation. [ABSTRACT FROM AUTHOR]

• MC_CNN network to detect secondary arc extinction in AC power transmission lines. • Multivariate time series data of phase voltage, current, and neutral reactor signals as input. • Extracts features in the convolution layer instead of hand-crafted feature extraction. • Detecting secondary arc extinction and signaling auto-reclose logic for autoreclosing application. • Feature learning via MC CNN outperform other types of ANN models. Detecting and recognizing the secondary arc extinction is considered as a backbone of the single pole auto reclosing schemes. This paper proposes a multi-channel convolutional network to detect the secondary arc extinction during a long AC power transmission line arc fault. In contrast to the conventional methods, the proposed Machin learning-based method can automatically learn the features from the raw data given as input. Multivariate time series data of phase voltage, current and neutral reactor signals were used as the input to the proposed model. The features are extracted in the convolution layer instead of using hand-crafted feature extraction like most of the existing research. The softmax classifier is used to detect fault and secondary arc extinction. Matlab Simulink is used to simulate the test system and collect the dataset to evaluate the proposed neural network model. To verify the generalizability of the proposed architecture datasets were collected at various locations and different phase lines. The proposed algorithm was able to detect the secondary arc extinction with 98.26% accuracy and able to give signal for auto-recloser logic with the maximum window length of three cycles. The result shows that the proposed neural network architecture is accurate and robust to detect the secondary arc extinction further can be used as a signal for a single pole adaptive auto recloser scheme. [ABSTRACT FROM AUTHOR]

[Display omitted] • Conventional image processing techniques are limited to 2D feature extraction. • A new approach combining deep learning algorithms and 3D modeling. • Machine learning methods can extract 3D information from 2D images. • Characterization of very dense multiphase flows with extreme particle overlap. • Mean volumetric properties of multiphase flows from 2D images are retrieved. Measuring the morphological properties of complex multiphase systems is a crucial problem in many areas of science and industry and is particularly difficult in dense environments with limited optical access. This paper presents a new approach capable of extracting three-dimensional (3D) information from spherical particle systems based solely on two-dimensional (2D) projections of the system. Synthetic images of the system are generated using a stochastic geometrical model from a simulated 3D particle system with the same geometrical features as the studied system, which is projected into 2D images labeled with the appropriate 3D information. These images are then fed to a convolutional neural network (CNN) for training before being tested on synthetic and experimental images. Validation results show that this technique successfully predicts the mean features of the studied systems, even for dense environments with overlapping particles, with high computational efficiency. [ABSTRACT FROM AUTHOR]

Nowadays, single image super-resolution based on deep learning algorithms is reaching state-of-the-art performance. Nevertheless, most frameworks suffer from needing huge sample set and powerful computing hardware for training. Therefore, this paper presents Cascading Parallel-structure units through a Deep-Shallow (CPDS) CNN which provides satisfactory results not only with small sample set but also low computational cost with the goal of fully using the network design capabilities. In the proposed deep-shallow CNN, an approximated image is reconstructed via the lightweight shallow network, and the missing high-frequency details are retrieved by its complementary deep network. Regarding this, deep CNN is designed based on cascading parallel-structure units in the whole parts of the model to reduce the depth of the model and the number of network parameters. Besides, in the proposed cascading strategy, feature disappearance is restricted by concatenating and fusing features. Based on the parallel-structure of deep CNN, the initial features are firstly extracted using two parallel feature extractor units. Moreover, deep CNN applies a multi-scale fashion to detect and extract coarse-to-fine features simultaneously. Finally, the strategy of combining upsampling methods of deconvolution layer and pixelshuffle is adopted to ensure the high-resolution image reconstruction is accurately validated. Experimental results demonstrate the proposed end-to-end approach outperforms most state-of-the-art methods in terms of training specifications, evaluation metrics, and visual performance. [ABSTRACT FROM AUTHOR]

This paper proposes a simple and effective model applied for image-based malware classification using machine learning in which malware images (converted from malware binary files) are directly fed into the classifiers, i.e. k nearest neighbour (k -NN), support vector machine (SVM) and convolution neural networks (CNN). The proposed model does not use the normalized fixed-size square images (e.g. 64 × 64 pixels) or features extracted by image descriptor (e.g. GIST) for training classifiers as existing models do in the literature. Instead, the input images are normalized and horizontally sized down (the width of the image) to a lower dimension of 32 × 64, 16 × 64 or even 8 × 64 than square ones (e.g. 64 × 64 pixels) to reduce the complexity and training time of the model. It is based on the fact that the texture of the malware image is mainly vertically distributed as analysed in this paper. This finding is significant for training those devices which have limited computational resources such as IoT devices. The experiment was conducted on the Malimg, Malheur datasets which contains 9339 (25 malware families) and 3133 variant samples (24 malware families) using k -NN, SVM and CNN classifiers. The achieved results show that it is possible to reduce the dimension of the input images (i.e. 32 × 64, 16 × 64 or even 8 × 64) while still retaining the accuracy of classification as the same as the accuracy obtained by classifier feeding by the fixed-size square image (i.e. 64 × 64 pixels). As a result, training time of the propose model reduces by a half, a quarter, and one-eighth compared to training time taken by the same machine learning-based classifier (i.e. k -NN, SVM and CNN) feeding by fixed-sized square images, i.e. 64 × 64, respectively. [ABSTRACT FROM AUTHOR]

Abstract Due to the rapid development of the information technology and wide use of the Internet, Information is easily to be obtained in the form of digital format. Digital content can be freely printed into documents since the convenience and accessibility of the printers. On the other hand, printed documents can be illegally manipulated by some criminal issues such as: forged documents, counterfeit currency, copyright infringement, and so on. Therefore, how to develop an efficient and appropriate safety testing tool to identify the source of printed documents is an important task in the meantime. Currently, the forensic system using the statistical methods and support vector machine technology has been able to identify the source printer for the text and the image documents. Such an approach belongs to the category of shallow machine learning with human interaction during the stages of feature extraction, feature selection and data pre-processing. In this paper, a deep learning system to solve the complex image classification problem is developed by Convolutional Neural Networks (CNNs) of deep learning which can learn the features automatically. Systematic experiments have been performed for both systems. For microscopic documents, feature based SVM system outperforms the deep learning system with limited gap. For scanned documents, both system can achieve equally well with high accuracy. Both systems should be constantly evaluated and compared for the best interest in universal utilization. Highlights • Deep learning can learn the texture features automatically. • For microscopic documents, feature based SVM classification still performs better. • For scanned documents, both system perform equally well with high accuracy. [ABSTRACT FROM AUTHOR]

Usage of depth data in various computer vision applications gains popularity with cheap depth sensors available on the market. In this study, a method that utilizes Deep Learning is proposed for single human actions and dyadic actions. Depth data of the actions are used to construct a three-dimensional (3D) template. These templates are rotated in different directions and two-dimensional (2D) views from different angles are stored. Acquired 2D images are used in deep feature extraction. AlexNet (Krizhevsky, Sutskever, & Hinton, 2012) pre-trained convolutional neural network is used for deep feature extraction. For all viewpoints, deep features are extracted and concatenated. After that random forest classifier is used for action recognition. The contributions of this paper are as follows. First, a 3D isosurface models is constructed to figure actions sequence. Second, we project 3D shapes into a 2D feature space by taking its snapshots from different views and giving them to a pre-trained CNN as input for feature extraction. We explore more information about the actions by extracting features from different layers of the deep neural network. We compare the results of the deep features extracted from different layers of the pre-trained CNN. Various classifiers are trained with extracted deep features. Because of the complex structure of the 3D shapes, there are a limited number of feature extraction methods. Our method obtains close to state-of-the-art performance on various single and two-person action datasets. [ABSTRACT FROM AUTHOR]

• Diabetic retinopathy (DR) detection from Fundus Images has been proposed. • Parallel CNN with fewer parameters and layers for distinctive feature extraction. • Extreme Learning Machine (ELM) technique was utilized for the DR classification. • The model demonstrated robustness for different types of datasets. • Accuracy upto 97.27 % was achieved and outscored state-of-the-art models. Diabetic retinopathy (DR) is an incurable retinal condition caused by excessive blood sugar that, if left untreated, can result in even blindness. A novel automated technique for DR detection has been proposed in this paper. To accentuate the lesions, the fundus images (FIs) were preprocessed using Contrast Limited Adaptive Histogram Equalization (CLAHE). A parallel convolutional neural network (PCNN) was employed for feature extraction and then the extreme learning machine (ELM) technique was utilized for the DR classification. In comparison to the similar CNN structure, the PCNN design uses fewer parameters and layers, which minimizes the time required to extract distinctive features. The effectiveness of the technique was evaluated on two datasets (Kaggle DR 2015 competition (Dataset 1; 34,984 FIs) and APTOS 2019 (3,662 FIs)), and the results are promising. For the two datasets mentioned, the proposed technique attained accuracies of 91.78 % and 97.27 % respectively. However, one of the study's subsidiary discoveries was that the proposed framework demonstrated stability for both larger and smaller datasets, as well as for balanced and imbalanced datasets. Furthermore, in terms of classifier performance metrics, model parameters and layers, and prediction time, the suggested approach outscored existing state-of-the-art models, which would add significant benefit for the medical practitioners in accurately identifying the DR. [ABSTRACT FROM AUTHOR]

Structural damage detection has been an interdisciplinary area of interest for various engineering fields. While the available damage detection methods have been in the process of adapting machine learning concepts, most machine learning based methods extract “hand-crafted” features which are fixed and manually selected in advance. Their performance varies significantly among various patterns of data depending on the particular structure under analysis. Convolutional neural networks (CNNs), on the other hand, can fuse and simultaneously optimize two major sets of an assessment task (feature extraction and classification) into a single learning block during the training phase. This ability not only provides an improved classification performance but also yields a superior computational efficiency. 1D CNNs have recently achieved state-of-the-art performance in vibration-based structural damage detection; however, it has been reported that the training of the CNNs requires significant amount of measurements especially in large structures. In order to overcome this limitation, this paper presents an enhanced CNN-based approach that requires only two measurement sets regardless of the size of the structure. This approach is verified using the experimental data of the Phase II benchmark problem of structural health monitoring which had been introduced by IASC-ASCE Structural Health Monitoring Task Group. As a result, it is shown that the enhanced CNN-based approach successfully estimated the actual amount of damage for the nine damage scenarios of the benchmark study. [ABSTRACT FROM AUTHOR]

Feature extraction plays a vital role in intelligent fault diagnosis of mechanical system. Nevertheless, traditional feature extraction methods suffer from three problems, which are (1) the requirements of domain expertise and prior knowledge, (2) the sensitive to the changes of mechanical system and (3) the limitations of mining new features. It is attractive and meaningful to investigate an automatic feature extraction method, which can adaptively learn features from raw data and discover new fault-sensitive features. Deep learning has been widely used in image analysis and speech recognition with great success. The key advantage of this method lies into the ability of mining representative information and sensitive features from raw data. However, the application of deep learning in feature leaning for mechanical diagnosis is still few, and limited studies have been carried out to compare the effectiveness of feature leaning with various data types. This paper will focus on developing a convolutional neural network (CNN) to learn features directly from frequency data of vibration signals and testing the different performance of feature learning from raw data, frequency spectrum and combined time-frequency data. Manual features from time domain, frequency domain and wavelet domain as well as three common intelligent methods are used as comparisons. The effectiveness of the proposed method is validated through PHM 2009 gearbox challenge data and a planetary gearbox test rig. The results demonstrate that the proposed method is able to learn features adaptively from frequency data and achieve higher diagnosis accuracy than other comparative methods. [ABSTRACT FROM AUTHOR]

Multimodal recognition of affective states is a difficult problem, unless the recording conditions are carefully controlled. For recognition “in the wild”, large variances in face pose and illumination, cluttered backgrounds, occlusions, audio and video noise, as well as issues with subtle cues of expression are some of the issues to target. In this paper, we describe a multimodal approach for video-based emotion recognition in the wild. We propose using summarizing functionals of complementary visual descriptors for video modeling. These features include deep convolutional neural network (CNN) based features obtained via transfer learning, for which we illustrate the importance of flexible registration and fine-tuning. Our approach combines audio and visual features with least squares regression based classifiers and weighted score level fusion. We report state-of-the-art results on the EmotiW Challenge for “in the wild” facial expression recognition. Our approach scales to other problems, and ranked top in the ChaLearn-LAP First Impressions Challenge 2016 from video clips collected in the wild. [ABSTRACT FROM AUTHOR]

Machine learning-based methods have shown great application prospects on the capacity estimation of lithium-ion battery. However, most of extant approaches require complete charging and discharging profiles that rarely occurs in practical applications to extract health features matched by the cycling capacity. To address this limitation, this paper directly intercepted partial charging voltage segments from raw charging curves, evading complicated manual features extraction. A progressive training framework is proposed and integrated with the deep convolutional neural network to build the online capacity estimator. Three battery aging datasets involving 28 cells cyclic aging data are exploited for model training and validation. Thereinto, the source dataset and target dataset-1 are used to train the model parameters progressively, and the target dataset-2 is applied to validate model performance. Experimental results show that the deep convolutional neural network coupled with progressive training framework can significantly enhance the capacity estimation accuracy and reduce retraining time compared with the ordinary convolutional neural network. The comparison results of different machine learning algorithms demonstrate that the proposed method possesses high estimation precision with a maximum relative error of only 3.4% on the target dataset, which can be easily extended to different types of battery capacity estimation. • The configuration of DCNN model is meticulously devised for capacity estimation. • Precise capacity estimation can be attained only using partial voltage segments. • A progressive training framework is proposed to improve DCNN model performance. • Capacity estimation does not require manual features extraction. [ABSTRACT FROM AUTHOR]

• An identification method for the subgrade settlement based on vehicle vibration signals and machine learning was proposed. • Vehicle vibration response law of high-speed railway caused by subgrade settlement was analyzed. • Realized the successful identification of subgrade settlement by using a combination of feature extraction, support vector machine, and convolution neural network. • Compared the CNN-SVM model with PSO-SVM in the identification results of subgrade settlement. High-frequency train loads and complex environmental factors over a long period will inevitably cause subgrade settlement, an issue that adversely affects line smoothness. The existing monitoring methods are inefficient, manpower consuming and small in scope. This paper takes the subgrade settlement disease of the double-block ballastless track as the research object and proposes a subgrade settlement identification method based on vehicle vibration signals. In addition, it establishes a vehicle-track-subgrade vertical coupling model that combines feature extraction with algorithms such as support vector machines (SVMs) and convolutional neural networks (CNNs) to accurately identify subgrade settlement diseases on ballastless tracks. The results show: I) the subgrade settlement is far more impactful than fastener failure and debonding of the supporting layer on the vehicle vibration, with the main frequency band concentrated in 0 to 10 Hz, and the vertical vibration acceleration of the vehicle body and bogie can be used as a sensitive feature to identify ballastless track subgrade settlement, II) both the particle swarm optimization (PSO)-SVM and CNN-SVM algorithms can effectively determine the existence of subgrade settlement and the values of the settlement wavelength and amplitude. Both algorithms run a 100 % identification of subgrade settlement, with the lowest identification rates of settlement wavelength being 97.56 % and 100 %, respectively and the lowest identification rates of amplitude being 84.78 % and 97.56 %, respectively. In general, the overall accuracy of the CNN-SVM always outperforms PSO-SVM, and III) changes to the vehicle passing speed will create a more obvious interference effect on the identification of the settlement amplitude. For the PSO-SVM algorithm, the identification rate for the amplitude has a difference of up to 8.70 %, while this is 2.44 % for the CNN-SVM algorithm, indicating the CNN-SVM algorithm is more robust than the PSO-SVM algorithm. [ABSTRACT FROM AUTHOR]

Finding efficient means of quantitatively describing material microstructure is a critical step towards harnessing data-centric machine learning approaches to understanding and predicting processing–microstructure–property relationships. Current quantitative descriptors of microstructure tend to consider only specific, narrow features such as grain size or phase fractions, but these metrics discard vast amounts of information. Since the gain in traction of machine learning and computer vision, more abstract methods for describing image data in a concise and quantitative manner have become available, but have yet to be fully exploited within materials science. The main aim of this paper is to investigate some of these methods as tools for constructing compressed numerical descriptions of microstructural image data, which are here referred to as "microstructural fingerprints". A statistical framework is developed to combine some of these methods which includes some classical computer vision methods as special cases. The effectiveness of fingerprints, in this case, is assessed via a series of classification tasks, which take the fingerprints as input along with some label to describe the microstructure, and aim to predict the class label. The classification tasks are a simple way to assess the information content of differently constructed fingerprints, using readily available data. However, the potential applications for such fingerprints, in theory, extend far beyond this, provided a suitably labelled training dataset. For example, fingerprints constructed from image data labelled with some processing parameters or mechanical properties could be fed into regression-based tasks to predict properties from microstructure or predict the processing parameters required for a target microstructure. The ultimate purpose is to rapidly fingerprint sample images in the context of various high-throughput design/make/test scenarios. Such fingerprints would enable, for example, quantification of the disparity between microstructures for quality control, classifying microstructures, predicting materials properties from image data and identifying potential processing routes to engineer new materials with specific properties. Here, the approach is applied to two distinct datasets to illustrate various aspects of the fingerprints and some recommendations are made based on the findings. In particular, methods that leverage transfer learning with convolutional neural networks (CNNs), pretrained on the ImageNet dataset, are generally shown to outperform other methods. Additionally, dimensionality reduction of these CNN-based fingerprints is shown to have negligible impact on classification accuracy for the supervised learning approaches considered. In situations where there is a large dataset with only a handful of images labelled, graph-based label propagation to unlabelled data is shown to be favourable over discarding unlabelled data and performing supervised learning. In particular, label propagation by Poisson learning is shown to be highly effective at low label rates. The supplementary code is available on GitHub (White and Tarakanov, 2021). [Display omitted] • Label propagation can outperform supervised learning on a reduced dataset. • Higher-order statistics on the feature set can increase classification accuracy. • Dimension reduction can boost classification speed without compromising accuracy. • Transfer learning works, but networks pretrained on microstructure are desirable. [ABSTRACT FROM AUTHOR]

• The Patch Stem module can guide the attention of the convolutional neural network. • This paper provides a new idea to identify coals and gangues from color images. • This method helps convolutional neural networks to better characterize the row coals. • The method can solve the problem of "different objects have the same spectrum". • This method makes the convolutional neural network identify gangues more accurate. Accurate identification of coals and gangues from color images can reduce the environmental damage caused by the mining industry. Coals and gangues are prone to the problem of "different objects with the same spectrum" in color images. This problem makes it difficult for convolutional neural networks to focus on discriminative features of color images, which makes it difficult to identify coals and gangues. To more accurately identify coals and gangues from color images, a novel attention mechanism inspired by the human visual system is proposed and named Patch Attention. Patch Attention consists of a Patch Stem module and a patch channel attention module. The Patch Stem module reshapes the image into a sequence of patches and performs feature extraction using convolution to locate discriminative features. The patch channel attention module further improves the feature response of discriminative regions in the image patch so that subsequent convolutional layers focus on these regions. The recognition accuracy of the model improved with the Patch Stem module can be improved by 2.54%, and the patch channel attention module inserted in the Patch Stem module can increase the recognition accuracy of the model by 0.8%. Patch Attention enables the recognition accuracy of convolutional neural networks to reach up to 99.37%. The Patch Stem module and patch channel attention module can significantly improve the recognition accuracy of coals and gangues from color images with convolutional neural networks. [ABSTRACT FROM AUTHOR]

Zero-shot learning enables the recognition of classes not seen during training through the use of semantic information comprising a visual description of the class either in textual or attribute form. Despite the advances in the performance of zero-shot learning methods, most of the works do not explicitly exploit the correlation between the visual attributes of the image and their corresponding semantic attributes for learning discriminative visual features. In this paper, we introduce an attention-based strategy for deriving features from the image regions regarding the most prominent attributes of the image class. In particular, we train a Convolutional Neural Network (CNN) for image attribute prediction and use a gradient-weighted method for deriving the attention activation maps of the most salient image attributes. These maps are then incorporated into the feature extraction process of Zero-Shot Learning (ZSL) approaches for improving the discriminability of the features produced through the implicit inclusion of semantic information. For experimental validation, the performance of state-of-the-art ZSL methods was determined using features with and without the proposed attention model. Surprisingly, we discover that the proposed strategy degrades the performance of ZSL methods in classical ZSL datasets (AWA2), but it can significantly improve performance when using face datasets. Our experiments show that these results are a consequence of the interpretability of the dataset attributes, suggesting that existing ZSL datasets attributes are, in most cases, difficult to be identifiable in the image. Source code is available at https://github.com/CristianoPatricio/SGAM. • An attention-based method for improving the performance of zero-shot face recognition. • Derivation of Attribute Attention Maps (AAM) for the most salient image attributes. • Extensive visual analysis of the AAM obtained from both CelebA and ZSL datasets. [ABSTRACT FROM AUTHOR]

Real-time and accurate traffic state identification can provide reference for urban traffic control and guidance. Due to the randomness and complexity of traffic flow, it is difficult to identify the traffic state accurately. The existing researches mainly adopted traffic state feature vectors and machine learning to identify the traffic state. Few studies attempted to use the feature map and deep learning for traffic state recognition. Therefore, this paper proposes a traffic state recognition model based on the traffic state feature map and deep learning. The feature map is a chromatogram form of digital traffic state feature vector. And the deep learning has strong predictive performance in image identification. In the model, the road traffic state feature vector is extracted from the vehicle trajectory data, and the Gram Angle Field (GAF) is adopted to transform the feature vector into feature map. Then, the deep learning algorithm is utilized in traffic state offline classification and online recognition. In the offline training phase, the historical vehicle trajectory data and the DeepCluster algorithm are used to establish the mapping relationship between the feature maps and the traffic states. In the online identification stage, the real-time vehicle trajectory data and CoAtNet algorithm are utilized to identify the current traffic state. In experiments, the proposed model is compared with the Convolutional Neural Network (CNN), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN) models using the open vehicle trajectory data of Shanghai North Cross Channel. The results showed that the proposed model achieved the best performance with the accuracy of 92.06% and could provide support for road traffic control and guidance. • Proposing a traffic state online recognition framework based on traffic state feature map and deep learning. • Adopting Gram Angle Field to construct traffic state feature map. • Using vehicles' trajectory data and CoAtNet to identify the traffic state. [ABSTRACT FROM AUTHOR]

Accurate State of Charge (SOC) and State of Health (SOH) estimation is crucial to ensure safe and reliable operation of battery systems. Considering the intrinsic couplings between SOC and SOH, a joint estimation framework is preferred in real-life applications where batteries degrade over time. Yet, it faces a few challenges such as limited measurements of key parameters such as strain and temperature distributions, difficult extraction of suitable features for modeling, and uncertainties arising from both the measurements and models. To address these challenges, this paper first uses Fiber Bragg Grating (FBG) sensors to obtain more process related signals by attaching them to the cell surface to capture multi-point strain and temperature variation signals due to battery charging/discharging operations. Then a hybrid machine learning framework for joint estimation of SOC and capacity (a key indicator of SOH) is developed, which uses a convolutional neural network combined with the Gaussian Process Regression method to produce both mean and variance information of the state estimates, and the joint estimation accuracy is improved by automatic extraction of useful features from the enriched measurements assisted with FBG sensors. The test results verify that the accuracy and reliability of the SOC estimation can be significantly improved by updating the capacity estimation and utilizing the FBG measurements, achieving up to 85.58% error reduction and 42.7% reduction of the estimation standard deviation. • A hybrid machine learning framework is proposed for battery states estimation. • Using Fiber Bragg Grating sensing technology to enrich process related measurements. • Useful features are automatically extracted from the enriched measurement signals. • Updating capacity and using FBG measurements enhance the accuracy of SOC estimation. • Using FBG measurements reduces the uncertainty of SOC estimation. [ABSTRACT FROM AUTHOR]

• A novel hyperparameter optimization based deep learning model for automated BAA. • Proposed method achieved average sensitivity 98.45% and average specificity 97.78% for the DHA dataset. • It achieved average accuracy 98.30%, average F-score of 98.31% for the DHA dataset. • It achieved an improved ROC of 99.9909% under the training set of 40% and maximum ROC values. • Under varying NFs, the proposed technique has resulted in maximal performance results. Bone age assessment (BAA) is widely employed in therapeutic investigation of endocrinology issues in children. BAA is usually carried out by radiological investigation of the left hand. Since manual BAA is prone to observer variations, it is needed to design automated BAA approaches. Recently developed deep learning (DL) models have demonstrated fascinating outcomes in automatic BAA. This paper presents a novel hyperparameter optimization based deep learning model for automated bone age assessment and classification (HPTDL-BAAC). The proposed HPTDL-BAAC technique aims to predict the bone age and classify it into several stages using X-ray images. The proposed model involves data normalization for pre-processing. For feature extraction, a Regional Convolutional Neural Network (RCNN) based mask with SqueezeNet is employed as a baseline model. For boosting the performance of SqueezeNet method, swallow swarm optimization (SSO) algorithm is used for hyperparameter optimization. Finally, SoftMax classifier-based age prediction and weighted extreme learning machine (WELM) based stage classification models are applied to determine the proper bone age and class label. A wide range of simulations were performed on Digital Hand Atlas (DHA) Database and the outcomes are examined with respect to several measures. The experimental outcomes highlighted the supremacy of HPTDL-BAAC technique over the other existing techniques. [ABSTRACT FROM AUTHOR]

Recognizing and segmenting complex texture images such as materials is of great significance to industrial design and production. Due to the lack of sufficient training samples and fuzzy boundaries in material images, it is difficult to segment material images by using deep learning methods. In material images, the pixels of each phase have a high degree of similarity, so if partial pixels' features in each phase are learned, the whole phase can be recognized. In this paper, we propose a method based on deep learning for recognizing and segmenting material images with complex textures. Firstly, the simple linear iterative cluster(SLIC) algorithm is used to obtain different numbers of superpixels which are a group of pixels with similar texture features. Then we extract the largest inscribed rectangular block in each superpixel. Next, put these rectangular blocks into the classical convolutional neural network(CNN)—DenseNet to recognize them. To retain the key texture features and reduce redundant information, we increase the receptive field in the key layers of DenseNet. In addition, due to the uneven distribution of phases in the material images, we improve focal loss to fit the material image. We make extensive comparative and ablation experiments to confirm the effectiveness of our method. [Display omitted] • Our method provides a new idea to solve the problem of small samples in material image segmentation. • Our method has strong feature extraction ability by adding feature enhance block (FE block) and our upsampling method of transition layer. • Our model has strong generalization ability, which saves a lot of time and cost. [ABSTRACT FROM AUTHOR]

Recently the most infectious disease is the novel Coronavirus disease (COVID 19) creates a devastating effect on public health in more than 200 countries in the world. Since the detection of COVID19 using reverse transcription-polymerase chain reaction (RT-PCR) is time-consuming and error-prone, the alternative solution of detection is Computed Tomography (CT) images. In this paper, Contrast Limited Histogram Equalization (CLAHE) was applied to CT images as a preprocessing step for enhancing the quality of the images. After that, we developed a novel Convolutional Neural Network (CNN) model that extracted 100 prominent features from a total of 2482 CT scan images. These extracted features were then deployed to various machine learning algorithms — Gaussian Naive Bayes (GNB), Support Vector Machine (SVM), Decision Tree (DT), Logistic Regression (LR), and Random Forest (RF). Finally, we proposed an ensemble model for the COVID19 CT image classification. We also showed various performance comparisons with the state-of-art methods. Our proposed model outperforms the state-of-art models and achieved an accuracy, precision, and recall score of 99.73%, 99.46%, and 100%, respectively. • Achieved an accuracy of 99.73% and a recall of 100%. • Enhance the quality of the CT scan images using CLAHE. • Build a novel CNN for extracting the most relevant features from the CT scan images. • Develop a soft voting ensemble learning model for improving the performance. [ABSTRACT FROM AUTHOR]

Coming with the emerging of depth sensors link Microsoft Kinect, human hand gesture recognition has received ever increasing research interests recently. A successful gesture recognition system has usually heavily relied on having a good feature representation of data, which is expected to be task-dependent as well as coping with the challenges and opportunities induced by depth sensor. In this paper, a feature learning approach based on sparse auto-encoder (SAE) and principle component analysis is proposed for recognizing human actions, i.e. finger-spelling or sign language, for RGB-D inputs. The proposed model of feature learning is consisted of two components: First, features are learned respectively from the RGB and depth channels, using sparse auto-encoder with convolutional neural networks. Second, the learned features from both channels is concatenated and fed into a multiple layer PCA to get the final feature. Experimental results on American sign language (ASL) dataset demonstrate that the proposed feature learning model is significantly effective, which improves the recognition rate from 75% to 99.05% and outperforms the state-of-the-art. [ABSTRACT FROM AUTHOR]

This paper proposes a sparse representation layer in the feature extraction stage of a convolutional neural network (CNN). Our goal is to add sparse transforms to a target network to improve its performance without introducing an extra calculation burden. First, the proposed method was achieved by inserting the sparse representation layers into a target network's shallow layers, and the network was trained end-to-end using a supervised learning algorithm. Second, In the forward pass the network captured the features through the convolutional layers and sparse representation layers accomplished with wavelet and shearlet transforms. Thirdly, in the backward pass the weights of the learned kernels of the network were updated through a back-propagated error, while the sparse representation layers were fixed and did not require updating. The proposed method was verified on five datasets with the task of image classification: FOOD-101, CIFAR10/100, DTD, Brodatz and ImageNet. The experimental results show that the proposed method leads to higher recognition accuracy in image classification, and the additional computational cost is relatively small compared to the baseline CNN model. • This paper proposes the hybrid of a CNN model and sparse representation. • We combine the CNN with predefined filters of wavelet and shearlet transform. • The paper theoretically explains the work procedure of the proposed network model. [ABSTRACT FROM AUTHOR]

Phase-sensitive optical time domain reflectometer (Φ-OTDR) can detect the occurrence of external vibration, but lacks of ability to classify multiple similar events. With the help of machine learning method, the events can be basically classified. An event recognition method based on deep learning and traditional classifier is proposed in this paper to further improve the classification accuracy. Firstly, the temporal-spatial data matrix from Φ-OTDR is directly sent to convolutional neural network (CNN) and the data features can be obtained automatically. Then these data features are transmitted to a traditional classifier for further classification. The traditional classifier can find a better classification hyperplane to improve the classification accuracy. In this paper, the most suitable classifier is support vector machine (SVM). As CNN works as a black box, T-distributed stochastic neighbor embedding (T-SNE) method and gradient-weighted class activation mapping (Grad-CAM) method are applied to visualize CNN's working process, which illustrates the correctness of feature extraction. Experimental results based on 11,997 images under eight event categories show that the classification accuracy can be improved by the traditional classifier. The CNN+SVM strategy performs the best and reaches 94.17 % accuracy, which offers 2.04 % improvement compared with using CNN alone. [ABSTRACT FROM AUTHOR]

Fault detection is a fundamental and important research topic in automatic seismic interpretation since the geometry of faults usually reveals the accumulation and migration of geological resource. In the communities of machine learning, many methods convert the fault detection task into a patch-based binary classification problem to solve. Because there are only a few faults in seismic data, the related classification task is obviously imbalanced, i.e., non-fault samples heavily outnumber fault samples. On considering that the existing CNN models just transform the problem into a balanced two-class task via undersampling, in this paper we attempt to construct CNNs directly from the imbalanced patch data with the assistance of transfer learning. In particular, we use focal loss to robustly learn fault feature representations for both fault and non-fault patches. With the aim to make full use of available data, we adopt transfer learning to enhance the recognition ability of relatively complex patches. The experiments conducted with both synthetic and real seismic data (some slices of the Netherland offshore F3 block in the North Sea) show that the novel method performs very well in detecting faults in seismic data. In supplemental material, the features extracted by the CNNs are visualized for both fault and non-fault patches to get some insights for their good performance. • We apply CNN with the focal loss to detect seismic faults with imbalanced patches. • Transfer learning strategy is used to fine-tune the CNN to classify complex patches. • We consider more types of faults in generated data to improve the CNN' robustness. • The CNN' predicted results are post-processed by some methods to extract fault lines. [ABSTRACT FROM AUTHOR]

• We propose a novel multi-view DCNN-based approach which can take advantage of not only multi-view images but also the relationships between them. • The proposed networks learn the integrated features of multi-view fundus images for DR detection. • We introduce the attention mechanisms for mining the relationship between different views. • In order to boost the ability to capture the tiny lesion features in the retinal images, we combine the idea of attention mechanisms with the channel dimension. Diabetic retinopathy (DR) detection has attracted much attention recently, and the deep learning algorithms have gained traction in this area. At present, DR screening by deep learning algorithms is often based on single-view fundus images, which usually leads to an unsatisfactory accuracy of DR grading due to the incomplete lesion features. In this paper, we proposed a novel diabetic retinopathy detection convolutional network for automatic DR detection by integrating multi-view fundus images. Compared to existing single-view DCNN-based DR detection methods, the proposed method has the following advantages. First, our method fully utilizes the lesion features from the retina with a field-of-view around 120 ∘ − 150 ∘. Second, by introducing the attention mechanisms, more attention will be paid on the influential view and the performance can be improved. Besides, we also assign large weights to important channels in the network for effective feature extraction. Experiments are conducted on our collected multi-view DR dataset contained 15,468 images, in which each eye sample provides four-view images. The experimental results indicate that using multi-view images is suitable for automatic DR detection and our proposed method is superior to other benchmarking methods. [ABSTRACT FROM AUTHOR]

• Optimal feature extraction from ultrasonic waveform essential for machine learning. • Convolutional neural networks pre-trained on auxiliary task. • Principal component analysis applied to CNN extracted features. • CNN feature extraction produced more informative features. • CNN method best for 65% of process monitoring tasks evaluated. Ultrasonic sensors are a low-cost and in-line technique and can be combined with machine learning for industrial process monitoring. However, training accurate machine learning models for process monitoring using sensor data is dependant on the feature selection methodology. This paper compares a convolutional feature extraction method to a traditional, coarse feature engineering approach. The convolutional method uses filter weights pre-trained on an auxiliary task to classify ultrasonic waveform dataset membership using previously obtained sensor data. The filter weights are used to extract features from the ultrasonic waveform. Principal component analysis is then applied to produce five principal components to be input into long short-term memory neural networks. The two approaches are compared on fermentation, mixing and cleaning datasets monitored using ultrasonic sensors. Overall, the convolutional feature method produced more informative waveform features than the coarse feature engineering approach, achieving higher model accuracy for datasets requiring substantial waveform information and for 65% of tasks overall. Multi-task learning also improved feature trajectory learning but led to reduced model accuracy for data points far from the classification decision boundaries. This can be overcome by further optimisation of neural network hyperparameters, though at increased model development time. Once trained, the convolutional feature extraction approach is a fast and convenient way of producing high quality features from ultrasonic waveforms using convolutional neural networks with little training data. [Display omitted] [ABSTRACT FROM AUTHOR]

Intrusion detection systems (IDS) differentiate the malicious entries from the legitimate entries in network traffic data and helps in securing the networks. Deep learning algorithms have been greatly employed in the network security field for large scale data in modern cyberspace networks because of their ability to learn the deeply integrated features. However, learning both space and time aspects of system information are very challenging for any individual deep knowledge model. While Convolutional Neural Networks (CNN) effectively acquires the spatial aspects, the Long Short-Term Memory (LSTM) neural networks perform better for temporal features. Integrating the benefits of these models has the potential for improving the large scale IDS. In this paper, a high accurate IDS model is proposed by using a unified model of Optimized CNN (OCNN) and Hierarchical Multi-scale LSTM (HMLSTM) for effective extraction and learning of spatial–temporal features. The proposed IDS model performs the pre-processing, feature extraction through network training and network testing and final classification. In the OCNN–HMLSTM model, the Lion Swarm Optimization (LSO) is used to tune the hyper-parameters of CNN for the optimal configuration of learning spatial features. The HMLSTM learns the hierarchical relationships between the different features and extracts the time features. Lastly, the unified IDS approach utilizes the extracted spatial–temporal features for categorizing the network data. Tests are performed over public IDS datasets namely NSL-KDD, ISCX-IDS and UNSWNB15. Assessing the performance of OCNN–HMLSTM​ against the contemporary IDS methods, the proposed model performs better intrusion detection with high accuracy of above 90% with less false values and better classification coefficients. [ABSTRACT FROM AUTHOR]

Time series classification (TSC) is the problem of categorizing time series data by using machine learning techniques. Its applications vary from cybersecurity and health care to remote sensing and human activity recognition. In this paper, we propose a novel multi-process collaborative architecture for TSC. The propositioned method amalgamates multi-head convolutional neural networks and capsule mechanism. In addition to the discovery of the temporal relationship within time series data, our approach derives better feature extraction with different scaled capsule routings and enhances representation learning. Unlike the original CapsNet, our proposed approach does not need to reconstruct to increase the accuracy of the model. We examine our proposed method through a set of experiments running on the domain-agnostic TSC benchmark datasets from the UCR Time Series Archive. The results show that, compared to a number of recently developed and currently used algorithms, we achieve 36 best accuracies out of 128 datasets. The accuracy analysis of the proposed approach demonstrates its significance in TSC by offering very high classification confidence with the potential of making inroads into plentiful future applications. [ABSTRACT FROM AUTHOR]

In the context of natural language processing, accuracy of intention detection is the basis for subsequent research on human-machine speech interaction. However, the problem of ambiguity in word vectors reduces the accuracy of intent detection. Meantime, there is a disconnection between local features and global features as well, resulting in text feature extraction that cannot fully reflect semantic information. These issues are all barriers of intention detection. Therefore, this paper proposes an attention-based convolutional neural network for self-media data learning (called A-CNN) for marketing intention. We cascade the traditional CNN with the self-attention model in the Attention networks to form a new network structure called A-CNN, and put forward a fast feature extraction method based on skip-gram-based learning called FSLText, to represent the high-dimension word vectors in the A-CNN. On the premise of maintaining the advantages of the CNN, A-CNN can not only solve the problem of local and global features disconnection caused by the CNN pooling layer, but also avoid the increase of algorithm complexity. The Self-Attention mechanism in the Attention model can effectively optimize the weight of local features of the information in global features, and retain local features that are more useful for intention detection. A fast feature extraction method which is based on Skip-gram can retain the semantic and word order information of the text. The method is beneficial to the marketing intention detection. According to the experiment, our A-CNN, compared with traditional machine learning methods, can improve 12.32% accuracy. Contrast to the dual-channel CNN, the accuracy rate is improved by 9.68%, and compared with the ATT-CNN, it is improved by 9.97%. On the F1 score, the A-CNN can improve the F1 score by about 9.37% in comparison with the traditional machine learning methods, the accuracy rate is increased by 9.68% compared with the dual-channel CNN, and 9.68% in contrast with ATT-CNN. It illustrates that our A-CNN can effectively address semantic and feature selection for marketing intention detection. [ABSTRACT FROM AUTHOR]

• Data augmentation performed to create a new sample image by performing transformation like flipping, rotation etc., to solve the problem of overfitting created due to the model complexity of a larger number of parameters. • Extra convolutional layer added to extract more important features and increase the overall accuracy in the detection of the left ventricle. • Drop out layer introduced during the feature extraction process to enable the model to learn only the important features and drop the irrelevant features to fix the problem overfitting and increases generalizability. deep learning algorithms have not been successfully used for the left ventricle (LV) detection in echocardiographic images due to overfitting and vanishing gradient descent problem. This research aims to increase accuracy and improves the processing time of the left ventricle detection process by reducing the overfitting and vanishing gradient problem. the proposed system consists of an enhanced deep convolutional neural network with an extra convolutional layer, and dropout layer to solve the problem of overfitting and vanishing gradient. Data augmentation was used for increasing the accuracy of feature extraction for left ventricle detection. four pathological groups of datasets were used for training and evaluation of the model: heart failure without infarction, heart failure with infarction, and hypertrophy, and healthy. The proposed model provided an accuracy of 94% in left ventricle detection for all the groups compared to the other current systems. The results showed that the processing time was reduced from 0.45 s to 0.34 s in an average. the proposed system enhances accuracy and decreases processing time in the left ventricle detection. This paper solves the issues of overfitting of the data. [ABSTRACT FROM AUTHOR]

• A novel probabilistic PV power forecasting method is proposed based on QCNN. • Two-stage training strategy (TSTS) is proposed to solve the training problem of QCNN. • QCNN possesses an evidently superior effect than four state-of-the-art methods. • TSTS provides references for the training of other quantile deep learning models. Probabilistic forecasting is significant in coping with the strong uncertainty of photovoltaic (PV) power, which provides the occurrence scope and corresponding probability information of PV power for the decision of power systems. Convolutional neural networks (CNNs), one of the most advanced and widely used deep-learning methods, are regarded as a promising method for predicting PV power. This paper proposes a daily-ahead probabilistic PV power forecasting method based on an improved quantile CNN (QCNN). First, the QCNN constructs a feature extraction network based on CNNs to mine the deep features of the PV power influence factors. Thereafter, quantile regression (QR) is employed to generate the PV power probability distribution based on the extracted features. However, QR produces non-differentiable loss functions that significantly hinder the training of QCNNs. To address this problem, a two-stage training strategy is proposed in which a CNN is trained using a deterministic forecasting method, and QR is trained using a linear-programming method. This strategy ensures that the training of QCNNs can avoid the effects of the non-differentiable loss functions such that the QCNN can play a complete role in the prediction. The case of a PV plant in Australia demonstrates that QCNNs produce a considerably superior effect when compared to the quantile extreme learning machine, quantile echo state network, direct QR, and radial basis function neural network. The proposed method can be beneficial in terms of decision-making in power systems. [ABSTRACT FROM AUTHOR]

In recent years, the success and capabilities of embedded vision have showed up in embedded applications. The embedding of vision into electronic devices such as embedded medical applications is being driven by the availability of high-performance processors, integrating with deep learning algorithms, as well as advances in image processing technology. But, including image processing in embedded vision systems need huge amount of computational capabilities even to process a single image to detect an object and it's extremely challenging to implement in embedded systems. Implementing deep learning algorithms and testing it on a task specific data set could provide enhanced results. In this paper, an approach for enhancing image processing architecture using deep learning for embedded vision systems is proposed and analyzed. Implementing deep learning algorithms and testing it on embedded vision yielded effective results. [ABSTRACT FROM AUTHOR]