Search

• Early prediction for mammography. • Classification of mammograms using feature extracted using Hough transform. • Strategies for classification and feature extraction. • The classification accuracy is more by the use of SVM classifier. • Proposed method is effectively classify the abnormal classes of mammograms. Breast cancer is one of the significant health problems in the world. If these abnormalities in breast cancer are detected early there is a maximum chance for recovery. For this early prediction we can go for mammography. It is one of the most effective and commonly used method for detecting and screening breast cancer. This paper presents classification of mammograms using feature extracted using Hough transform. Hough transform is a two dimensional transform. It is used to isolate feature of particular shape in an image. Miniaturized scale characterization and masses are the two most vital markers of threat, and their mechanized identification is exceptionally important for early breast cancer diagnosis. Since masses are regularly undefined from the encompassing parenchymal, computerized mass location and arrangement is significantly additionally difficult. This paper talks about the strategies for classification and feature extraction. Here, Hough transform is used to detect features of mammograms image and it is classified using SVM. The classification accuracy is more by the use of SVM classifier. This method is tested on 95 mammograms images collected and classified using SVM. From the result it shows that the proposed method is effectively classify the abnormal classes of mammograms. [ABSTRACT FROM AUTHOR]

• Now a days the major oral health issues of man is crucial an area of research. • This paper traces classification, segmentation and pre-processing of dental images. • Histogram based adaptive approach is used in the radiographic X-ray 2d dental Image. • This is useful to distinguish the foreground teeth & the regions of background bones. In the present scenario the major oral health issues of man is crucial an area of research. The Data mining techniques, image processing, and Computational intelligence techniques are playing a vital role in biomedical research. Dental image processing helps to improve the early detection and classification of the diagnostic process to make accurate decisions. The radiographic 2d dental image is widely utilized for analytic thinking of several dental disorders. In this paper traces the complete steps such as classification and segmentation as well as pre-processing of dental images has been carried out. In the pre-processing, histogram based on adaptive approach is used to stretch the contrast and equalize the brightness throughout the radiographic X-ray 2d dental Image. This operation is useful to distinguish the foreground teeth and the regions of background bones. Separation of dental 2d images into regions corresponding to the objects is a fundamental step of segmentation. The hybrid graph cut segmentation is used to segment the oral cavity and its tissues. In this research deep learning based convolution neural network (CNN) has been used to process the dental image and shows promising outcomes with 97.07% accuracy. [ABSTRACT FROM AUTHOR]

One of the main challenges in BCI technologies is the problem of poor information transfer rate (ITR) of BCI systems and its effect on reducing the commands users can give. At the state of the art, the Linearly Constrained Minimum Variance (LCMV) beamformer is one of the most effective strategies that can provide satisfactory performance in the analysis of steady-state visual evoked potentials (SSVEP). When only a short-period SSVEP signal is available, a poor estimate of the covariance matrix used in the original LCMV method is obtained. Therefore, the main drawback of the conventional LCMV beamformer and its spatiotemporal form (LCMVst) is that they require a long-period SSVEP signal to achieve a reliable estimate for the beamformer weights. To cope with the mentioned problems, this paper presents two types of regularized estimators to achieve a robust estimate for the covariance matrix when only a short-period SSVEP signal is available. These two types, which are user-independent, include the convex combination (CC) and general linear combination (GLC) methods. In the next step, in order to provide suitable beamformer weights, the CC and GLC covariance estimators are combined with the original LCMVst beamformer. Furthermore, two new subject-based electrode selection algorithms, i.e., sequential forward electrode selection (SFES) algorithm and common spatial pattern (CSP) based technique, are applied to pick out the best collection of electrodes and to maximize the classification accuracy. The introduced methods were assessed using a four-class SSVEP dataset recorded on 20 subjects. Experimental results indicate that the proposed estimators optimally classify SSVEPs by estimating proper beamformer weights. The proposed LCMVst-CC and LCMVst-GLC beamformers improved the average classification accuracy by about 27.75% in comparison to the original LCMV-based beamformer at the 0.25-second segment length. Using SSVEP stimulations, we show that our proposed beamformers offer a better classification performance than the conventional LCMV-based spatiotemporal beamformer for all stimulation times. [Display omitted] • Two new adaptive and robust beamformers are proposed to classify the SSVEP responses. • A new sequential forward electrode selection (SFES) method is proposed in this paper. • The GLC and CC-based methods can significantly improve the classification accuracy. • The proposed beamformers improved the ITR value by analyzing short-time EEG trials. • The proposed methods outperform existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

• Sparsification process is explored for multi-source high-dimensional profile data. • Models with spatial geometric interpretations are established for anomaly monitoring and fault diagnosis. • Estimated weight vector of an optimal linear hyperplane model is interpretable. • Absolute weights represent feature importance and their signs represent direction influence on classification. Condition monitoring and fault diagnosis is important to ensure smooth machine operation, in which the use of multi-source high-dimensional profile data composed of multiple physical signals collected by multiple sensors for machine health monitoring and fault diagnosis has not been fully explored. In this paper, interpretable monitoring and diagnostic methods are proposed to analyze multi-source high-dimensional profile data and identify equipment operational status. Firstly, polynomials of segmenting profiles as well as a Lasso model are established to obtain sparse features. Then, for offline data with historical fault labels, an optimal linear hyperplane classifier is constructed to interpretably study the impact of different features on a decision. For online data with only normal samples, constructing an optimal hypersphere enables the measurement of operational status. Large-scale nuclear-grade hydraulic damper profile data are used to demonstrate how the proposed interpretable models work and show their physical interpretation for monitoring and fault diagnosis. [ABSTRACT FROM AUTHOR]

Although surface defect detection for roller bearings has received much attention in smart industrial manufacturing, most existing approaches employed the existing Convolution or Transformer models without modification to detect the defects. Such applications are undesirable because the classification models lack the necessary spatial relationship between multiple objects within an image. To develop a specialized model to detect bearing defects, we expand the applicability of the Swin Transformer. This paper presented a shunted-window Transformer (sSwin), which replaces the shifted-window mechanism with the shunted large–small windows. In sSwin, the self-attention heads can capture multi-scale token granularity features within the same self-attention layer. Also, a proposed local connection module enhances the boundary interaction of adjacent value tokens. As a result of the different-scale receptive fields of tokens within one attention layer, the model can bring significant performance gain on the defective detection tasks. The experiment demonstrates that sSwin-T outperforms the classical Swin-T by +2.8% AP b and +1.9% AP m , and Swinv2-T by +1.0% AP b and +1.1% AP m on COCO. It also surpasses them by +1.1% AP b and +3.4% AP b on the private bearing dataset with Cascade Mask-RCNN 3x. • Introducing a novel shunted swin transformer network for the purpose of bearing surface defect detections. • Developing a shunted window mechanism within a self-attention layer. • Capturing the coarse- and fine-grained defective regions simultaneously during self-attention computation. • Offering roust generalization to adapt to various measurements and conditions. [ABSTRACT FROM AUTHOR]

Heart sound analysis is a non-invasive and economical technique that can aid in diagnosing cardiovascular disease. A novel End-to-End heart sound classification method was proposed in this paper, in which a combination of multi-scale dense network and multi-head recurrent neural network technology was used. It can be used to diagnose congenital heart disease (CHD) without using the manual extraction of features. An F β score of 94.33% and an accuracy of 94.41% were achieved by the method on dataset A, which consisted of 1,000 individuals and 5,000 signals. Similarly, the widely used dataset B (Physio Net/CinC 2016 dataset), comprising 764 individuals and 3,240 signals, resulted in an F β score of 93.75% and an accuracy of 92.97%. The results show the proposed method had a significant potential to assist in diagnosing CHD. The SHAP algorithm which is a kind of Interpretable method was applied in this study to interpret the prediction results of model. It was shown that the model's prediction process is similar to a doctor's diagnosing mode. • Heart sound classification without manual feature extraction. • Multiscale densely connected network. • Exploring the interpretability of model classification results using SHAP. [ABSTRACT FROM AUTHOR]

[Display omitted] • An experimental setup was built up to collect the VILAES. • The wavelet scattering coefficients are selected as the features of the VILAES. • Relationship between scattering coefficients and leakage rate is determined. • Optimal wavelet scattering coefficients are extracted automatically. • A new multi-variable classification model for small leakage and small sample sizes. In order to achieve acoustic emission detection in valve internal leakage, it is essential to extract features, and establish an accurate mathematical model. Current valve internal leakage acoustic emission signal (VILAES) classification models mostly rely on human experience for selecting features, resulting in low accuracy for small leakage conditions. This paper combines the wavelet scattering transform (WST), Relief-F algorithm and AdaBoost.M1 algorithm, and proposes to use the optimal wavelet scattering coefficients as features to establish the VILAES classification model accurately for small leakage. Firstly, the first three order wavelet scattering coefficients of the VILAES are automatically extracted using WST and are transformed into one-dimensional features. Secondly, the Relief-F algorithm is employed to select the optimal feature subset. Finally, the optimal wavelet scattering coefficients and pressures are used as inputs to establish a classification model for VILAES and achieve an accuracy over 96.80% for small leakage. [ABSTRACT FROM AUTHOR]

[Display omitted] • The acquisition system uses ultrasonic sensors to detect vehicles and then creates a dataset of audio recordings of the engine idling. • The dataset of audio samples is analyzed by an unsupervised learning algorithm that uses psychoacoustic features to classify vehicles as those with gasoline or diesel engines. • A self-organizing map used for classification achieves an F1 score of 91.1%, which increases to 96.7% when vehicles with mechanical anomalies are excluded. Much research has been done in the field of classifying vehicles based on their fuel type. One of the many potential applications is to improve the quality of life in urban areas by separating vehicles based on their pollution level. Real-time classification and implementation of appropriate on-site IoT measurement devices is critical to developing a system that accurately identifies vehicle's fuel type without violating driver privacy. In this paper, a classification system based on psychoacoustic features extracted from sound recordings is investigated. Unsupervised learning was implemented as it is able to detect hidden connections within the input space without relying on labelling data. Our goal was to develop and explore a relatively fast classification system, focusing on a short acquisition time. A self-organizing map with a 10-dimensional input space was used and effective classification with five single-valued features was demonstrated. The analysis of the input space shows the difficulty and complexity of such an approach and leaves room for further improvement. [ABSTRACT FROM AUTHOR]

This paper investigates the sensitivity of statistical fingerprints to different phase resolution (PR) and amplitude bins (AB) sizes of partial discharge (PD) φ - q - n (phase-amplitude-number) patterns. In particular, this paper compares the capability of the ensemble neural network (ENN) and the single neural network (SNN) in recognizing and distinguishing different resolution sizes of φ - q - n discharge patterns. The training fingerprints for both the SNN and ENN comprise statistical fingerprints from different φ - q - n measurements. The result shows that there exists statistical distinction for different PR and AB sizes on some of the statistical fingerprints. Additionally, the ENN and SNN outputs change depending on training and testing with different PR and AB sizes. Furthermore, the ENN appears to be more sensitive in recognizing and discriminating the resolution changes when compared with the SNN. Finally, the results are assessed for practical implementation in the power industry and benefits to practitioners in the field are highlighted. [ABSTRACT FROM AUTHOR]

[Display omitted] • This paper shows how the problem of CM&FD of OOR wheels can be transformed into an ML problem. • A vehicle-track MBS model considering track irregularities and worn wheel profiles is built. • A DL model, OORNet, is proposed to realize CM&FD of our-of-round wheels. • The feasibility and superiority of OORNet are demonstrated by three other DL models. The problem of train wheel out-of-roundness (OOR) negatively affects both humans and the vehicle-track system, incl. reduced passenger comfort, rapid aging of vehicle/track components, increase in derailment risk, etc. It is therefore of interest to develop an on-board condition monitoring and fault diagnosis (CM&FD) technique for wheel OOR, which contributes not only to the maintenance decision-making of wheelsets but also to clarifying its triggering and evolution mechanisms. This paper first shows how to express the problem of CM&FD of our-of-round wheels as a machine learning problem. A deep learning model, OORNet, is then developed for CM&FD of out-of-round wheels. A vehicle-track multi-body dynamics model of a China railway high-speed (CRH) trailer is meanwhile built to produce a database consisting of vertical axlebox vibration accelerations caused by 2000 different wheel OOR curves. The simulated database is finally used to test the performance of OORNet, and its feasibility and superiority are verified. [ABSTRACT FROM AUTHOR]

The paper presents the construction and application of multiple classifiers to increase the accuracy of the fault detection module in the diagnostic task. The structure of the ensemble of classifiers is presented and the applied voting mechanisms explained. Methods of storing knowledge in the intelligent diagnostic systems are introduced and their taxonomy provided. Next, the selected algorithms implemented in the fault detection operation are briefly described. Problems with the practical implementation of the proposed solution are considered. The scheme is used to detect faults in the analog part of the MEMS accelerometer. The paper is concluded with the possible prospects for the proposed scheme. [ABSTRACT FROM AUTHOR]

The advancement of brain–computer interfaces (BCIs) has narrowed the gap between humans and computers, allowing intentional interaction by monitoring and translating brain signals in real time. Among BCI approaches, motor imagery electroencephalogram (MI-EEG) systems are popular due to their non-invasiveness, portability, and user-friendly operation without external stimuli. However, MI-EEG classification faces challenges from subject-specific variations in time, frequency, and spatial domains. To overcome this, the paper proposes a novel Hilbert–Huang transformation (HHT)-based method for subject-specific time–frequency-space pattern optimization in MI-EEG classification. The method utilizes a joint time–frequency pattern optimization module and a spatial pattern optimization module for EEG measurements. This efficient process identifies subject-specific dominant time–frequency components and extracts optimal spatial features. The optimized features are fed into a support vector machine (SVM) classifier, resulting in superior performance compared to standard baselines on three open-source datasets. The proposed method achieves 4.1% and 6.3% higher accuracy in 2-class and 4-class classification, respectively. Additionally, it demonstrates remarkable computational efficiency, requiring 70% less training time to achieve the optimal feature space. These improvements in classification accuracy and computational efficiency underscore the practical value of the proposed method for MI-BCI systems. • Efficient HHT-based method for precise subject-specific TF component identification. • Spatial EEG filtering with focus on dominant time-frequency components. • Outperforms baselines, achieving 4.1% and 6.3% higher accuracy on open datasets. [ABSTRACT FROM AUTHOR]

• Introduce a new approach that using spectrograms for feature extraction. • Newly propose a complete route of feature optimization to construct a high-quality dataset. • Introduce a new fusion classification model achieving a very high prediction accuracy. • Contribute to the animal welfare and promote the development of broiler farming practices. This paper presents a sound recognition method for white feather broilers using spectrogram features and a fusion classification model, with the goal of achieving accurate classification of white feather broilers sound signals and providing a reliable basis for monitoring their health. In the training part, after five steps of sound signal acquisition, pre-processing, feature extraction, feature optimization, and model training, a fusion classification model with strong reliability is constructed for practical application scenarios. In the testing part, the method is applied to a real farming scenario of white feather broilers, and the stability of the multi-classification models and the reliability of the fusion classification model are verified. The fusion classification model comprises Random Forest, K-nearest neighbor, and RBF-based SVM. Results from multiple tests showed that the highest classification accuracies achieved by the three multi-classification models were 100%, 86.67%, and 93.33%, respectively. The average prediction accuracy of the fusion classification model on multiple audio signals was 98.57%, the results effectively demonstrate the feasibility and practicality of the proposed method. [ABSTRACT FROM AUTHOR]

Abstract: Effective fault location classification and especially performance degradation assessment of a roller bearing have been the subject extensive research, which can reduce costs and the nonscheduled down time. In this paper, a new fault diagnosis method based on multiple features, kernel principal component analysis (KPCA) and particle swarm optimization-support vector machine (PSO-SVM) is put forward. First, traditional features of the vibration signals in time-domain and frequency-domain are calculated, and then two types of features referred to as singular values and AR model parameters based on ensemble empirical mode decomposition (EEMD) are introduced. After that, the original feature vectors are mapped into higher dimensional space and the kernel principal components are extracted as new feature vectors, which are used as inputs to PSO-SVM. The experimental results show that the new diagnosis approach proposed in this paper can identify not only the fault locations but also the performance degradation of the roller bearing. [Copyright &y& Elsevier]

Abstract: In this paper, a new portable instrument for the analysis of telecommunication signals is proposed, which is able to: (i) observe a wideband in which even multiple signals can be present at different carrier frequencies; (ii) divide the observed wideband in smaller sub-bands and detect the presence of a signal in each sub-band; and (iii) recognize the modulation type of the detected signals, among the most common digital schemes, without requiring any knowledge about their parameters. The proposed instrument, completed with the signal-dependant measurement routines, represents a considerable advancement of the research among the instruments for telecommunication systems, as it is capable of working autonomously on unknown signals. Such feature avoids both the intervention of a skilled operator in setting up the measurement task, and make it ready to be used in the next generation telecommunication scenarios, where a more flexible use of the radio spectrum and a dynamic allocation of users are foreseen. The hardware architecture is based on a PXI data acquisition system and a laptop PC processing unit while the control software has been developed in LabVIEW. The paper proposal has been validated using actual signals in several scenarios emulating real world communications. [Copyright &y& Elsevier]

• A nature-inspired white blood cells classification technique has been proposed. • OBBA has been proposed for image feature selection. • KNN, Logistic Regression, Random Forest & Decision Tree classifiers has been used. • OBBA select 11 features with an accuracy of 97.3%. The quantitative and differential analysis of leukocytes present in human body provide conducive hematological information to physicians for diagnosis of various infections and ailments. This paper proposes an Optimized Binary Bat algorithm, an enhanced version of the original Binary Bat Algorithm, for classification of different types of leukocytes. It is used for the first time in this field of application to the best of our knowledge. A set of features are extracted from images of WBCs and then the optimized algorithm is used to obtain a subset of those features which are essential and more relevant from the high-dimensional dataset. Similar to the original BBA, the optimized BBA is an evolutionary algorithm inspired by the echolocation technique used by bats for locating a prey or an object. OBBA aims to reduce the dimensionality of the dataset by determining the features which are most discriminative. The proposed algorithm is implemented using four different classifiers, K-nearest neighbors (KNN), Logistic Regression, Random Forest and Decision Tree, and their performance is compared. The proposed OBBA can be used in classification of WBCs with an average accuracy of 97.3% and help in analysis of numerous hematological conditions. The optimized BBA is also compared with other nature inspired algorithms including Optimized Crow Search algorithm and Optimized Cuttlefish algorithm, and the results indicate that the suggested algorithm is sufficiently fast and accurate to be used in hematological analysis. [ABSTRACT FROM AUTHOR]

Highlights • A block-based perceptual hash function is developed. • Size of images covered on the hard disk reduced by between 20 and 180 times. • CNN architecture is used as feature extractor to avoid manual feature extraction. • These features are used in various classification (SVM-KNN-ELM) algorithms. • A new hybrid PH-C-ELM method is proposed for liver image classification. Abstract Classification of liver masses is one of the hot topics in the literature. This paper proposes a hybrid method of using Convolutional Neural Network (CNN) and Discrete Wavelet Transform- Singular Value Decomposition (DWT-SVD) based perceptual hash function. The aim of the proposed method is to reduce the execution time of CNN architecture, space of liver images occupied on the hard disk and maintain the classification performance above an acceptable threshold. The proposed method has been designed for classifying malignant and benign masses from liver CT images. The most important features required for classification are achieved by the acquisition of salient features using Perceptual hash functions. Experimental evaluation was performed with 5-fold cross validation on a set of 200 CT images, 100 of benign tumors and 100 of malignant tumors. Results showed that the CNN features achieved high classification performance with different classifiers. However, experimental results show that CNN features achieved better classification performance with ELM, where ELM simulation results validated output data with success 97.3%. [ABSTRACT FROM AUTHOR]

Highlights • This study applies convolutional neural network to flower species classification. • It employs a pre-trained DCNN model for feature extraction. • Minimum Redundancy Maximum Relevance (mRMR) method is used to select features. • A support vector machine (SVM) classifier is employed to classify the flower species. • It achieved 96.39% and 95.70% accuracy performance for Flower17 and Flower102. Abstract Image-based automatic flower species classification is an important problem for the biologists who construct digital flower catalogs. A dozen of work about flower species recognition has been proposed so far based on traditional image processing routines. Nowadays, researchers apply the deep learning on various image-based object recognition tasks. In this paper, deep convolutional neural networks (DCNN) based hybrid method is applied to flower species classification. The proposed method initially employs a pre-trained DCNN model for feature extraction. To this end, two popular DCNN architectures namely, AlexNet and VGG16 models are adopted. For constructing efficient feature sets, the features from AlexNet and VGG16 models are then concatenated. Finally, a feature selection algorithm, the minimum Redundancy Maximum Relevance (mRMR) method, is used to select the more efficient features. A support vector machine (SVM) classifier with Radial Bases Function (RBF) kernel is employed to classify the flower species using the extracted features. Flower17 and Flower102 datasets which have a huge amount of category are used in the experimental works. Various experiments results show that we have achieved 96.39% and 95.70% accuracy performance for Flower17 and Flower102, respectively. The obtained results demonstrate the effectiveness of the proposed method, despite the relative simplicity of the approach. [ABSTRACT FROM AUTHOR]

• Time-resolved CFI image acquired. • The fast rise section and the slow decrease section of the ChlF kinetics curve were measured. • Characteristics were extracted and a ANN model was established. The objective of this study was to check whether different water and nitrogen treatments and, even the water-nitrogen coupling effect of plants could be correctly differentiated via chlorophyll a fluorescence image. We developed a classification method using the imaging analysis of chlorophyll a fluorescence induction based on Artificial Neural Network. The measurements were carried out on scheffera octophylla (Lour.) Harms , and the images were recorded at 690 nm with a high-resolution imaging device consisting of LEDs for an excitation at 460 nm and an Electron-Multiplying CCD camera. The effect of three different water and three different nitrogen treatments on the fluorescence parameters were obtained by hundreds of time-resolved fluorescence images. We used a Radial Basis Function neural network to model and test the sample data. The results showed that the different water and nitrogen statuses of plants were identified by the chlorophyll a fluorescence images and showed a high recognition accuracy. Compared with nitrogen, water had more of an influence on chlorophyll a fluorescence and was easier to identify. However, because the water and nitrogen restrict and promote each other, studying the coupling effect of water and nitrogen is necessary. Nine levels of water-nitrogen coupling plants were tested and classified. We discovered that a significant decrease on the classified accuracy was observed for the high nitrogen and low nitrogen treatments, while under a medium N-supply, the recognition rate was high. The method in this paper allowed plants to be classified under different water and nitrogen treatments, and has the potential to monitor the water and nitrogen coupling effect of plants in situ. [ABSTRACT FROM AUTHOR]

[Display omitted] • In this study, a 2D CNN layer is utilized that extracts spatial features and a dual-stream transformer block is added to provide global attention. • Through two-way routes, precise views of local and global features at the pixel level can be obtained. The increased self-attention information is distributed parallel along the semantic and pixel paths to train the model jointly. • The model's superiority is validated on three datasets for binary and multi-class classification. The obtained performance is notable and can be used for a real-time crack detection system. This paper addresses the problem of automatic crack identification within building structures. Cracks can cause structural damage to a building, necessitating prompt repairs to preserve lives. Detecting structural defects is a difficult process, especially in tall towers. Machine learning and deep learning approaches automate damage identification and crack detection. Machine learning-based methods train models with manually produced features that may contain errors. Deep convolutional neural networks (DCNN), on the other hand, enhance performance by extracting high-dimensional features. However, DCNN algorithms for crack identification lack a global focus on features that can compromise performance. This study inserts a dual-stream transformer module into a single 2D convolution neural network (CNN) layer. The dual-stream block contains a semantic and pixels path. Precise views of local and global features at the pixel level can be obtained through two paths. This increases self-attention information distributed parallel along the semantic and pixel paths to train the model. Further, the DSTNet (Dual Stream Transformer Network) superiority is validated on the three datasets and achieved notable performance. [ABSTRACT FROM AUTHOR]

• A UMS-FRCNN method for automatic detection and classification. • Improving Classification Accuracy with Convolutional Region Proposal Networks. • Detect small textured regions using a region proposal component. • The network is trained end-to-end through a fully joint training method. In traditional industrial production, the classification of cork mainly adopts manual operation. Due to round cork blocks with complex textures, manual sorting is time-consuming and often leads to inefficiency. It will lead to inefficiency and a high error rate. Therefore, an automatic cork sorting system is urgently required to replace manual operation. In this paper, an advanced deep-learning method based on machine vision is adopted to detect the surface of corks and classify them automatically. To robustly detect the texture of cork profile, a Multiple Scale Faster R-CNN based on the upper and lower layers (UMS-FRCNN) approach is proposed. The network consists of a region proposal component and the region-of-interest (RoI) detection component. There are two main contributions in our proposed network, which play a significant role in achieving state-of-the-art performance in cork classification. Firstly, the multi-scale information is grouped in two aspects: region suggestion and RoI detection, to detect on small cork texture regions. Secondly, the convolution Region Proposal Network (RPN) is used to generate the candidate regions of the UMS-FRCNN model to improve the classification accuracy. The experimental results show that our method not only can accurately classify the corks but also has better performance than the Faster RCNN. Compared with other methods, our method has the advantages of high precision, less time-consuming, and no dependence labeling. Our work has been successfully utilized in a cork production plant. [ABSTRACT FROM AUTHOR]

• This paper proposes a novel algorithm for screening EIMFs. • This paper applies EIMFs to signal denoising and feature matrix construction. • This algorithm improves the problem of TVF-EMD to decompose too many imfs. • This algorithm expands the application of TVF-EMD in the field of fault diagnosis. This paper proposes a new narrow-band filtering algorithm to improve the problem of TVF-EMD algorithm decomposing too many narrow-bands. The algorithm uses the energy estimation model of IMFs combined with the energy of noise in each imf and the signal complexity evaluation standard to obtain the effectiveness operator that measures the signal content of each imf, and selects the eimf with a large effectiveness operator as the EIMFs. In this paper, three groups of rotating machine data are used for experiments. The classification accuracy of denoising signals can reach 99.98% when the effectiveness operator is accumulated to 0.9999, and the classification accuracy of the EIMFs feature matrix can reach 97.83%, which are higher than the original data control group. The algorithm only needs to deal with the advantages of EIMFs, which significantly improves the classification accuracy and iteration speed of the classifier. [ABSTRACT FROM AUTHOR]

• Random Forest and Multi-Layer Perceptron (MLP) classifier. • The breast cancer features are generated using radial based function network. • Experiments with the Random Forest model and MLP model. Breast cancer (BC) is the most commonly found disease among women all over the world. The early diagnosis of breast cancer can potentially reduce the mortality rate and increase the chances of a successful treatment. Paper focuses on proposing a methodology to conduct early diagnosis of breast cancer using the Internet of Things and Machine Learning. The main objective of the paper is to explore the machine learning techniques in predicting breast cancer with IoT devices.Proposed classifier resulted in 98%, 97%, 96% and 98% of precision, recall, F_Measure and accuracy, respectively. The minimum error rate for the classifier have also been determined and found to be 34.21%, 45.82%8, 64.47% of Mean Absolute Error (MAR), Root Mean Square Error (RMSE) and Relative Absolute Error (RAE), respectively. It was evident through the obtained results that the MLP classifier yields a higher accuracy with a minimum error rate when compared to LR and RF. [ABSTRACT FROM AUTHOR]

The paper presents practical application of an electronic nose technique to fast and efficient discrimination of the samples of different fungi species such as: Penicillium chrysogenum, Cladosporium herbarum, Rhizopus oryzae , Alternaria alternata. Two prototypes of electronic nose instrument were utilized for investigation of discrimination capability with respect to odour profile of these fungi: the first prototype was based on MOS-type chemical sensors, the second one was based on gas chromatography technique. The fungi samples were prepared as the aqueous suspensions, the headspace of which was subjected to analysis. The data were analysed using the multidimensional methods: PCA, LDA and k-NN. The obtained results confirmed legitimacy of application of the electronic nose technique to identification and discrimination of fungi species. In case of the gas chromatography-based electronic nose prototype correct classification of the fungi species was at the level of 80–100% depending on the classification method employed, in case of the electronic nose prototype utilizing the MOS-type sensors correct classification was at the level of 55–100%. [ABSTRACT FROM AUTHOR]

Power cables, the most critical component of the power system, must be extremely reliable in order to avoid revenue losses due to premature failure. The dielectric properties of cable insulation may deteriorate due to ageing phenomena, which may have a negative impact on the polymer materials used for insulation. Therefore, the early detection of such depletion, and the severity of degradation while the equipment is in operation, aids in the avoidance of a total failure. Partial discharge (PD) detection and analysis have been adopted as a predictive test to characterize and assess the state of electric cables in advance. This review provides an in-depth discussion of the reactions that occur in the insulation system of the cables. Moreover, this paper presents a comprehensive review of the state-of-the art of various PD detection techniques regarding sensor types. The drawbacks and challenges of different PD measurement techniques have been elaborated. Following that, the numerous PD localization methods are discussed, as well as the necessity of computational intelligence approaches and their pros and cons. Last but not least, the authors provide a deep insight into the theoretical and practical implications of deep learning in PD localization, as well as recommendations for future research directions. This review will provide valuable insights and act as a starting point for researchers to lead the development of more efficient approaches for diagnosing PD in the cable. [ABSTRACT FROM AUTHOR]

Closed-circuit television (CCTV) is being widely adopted in water pipeline inspection. The inspector needs to spend a long time to watch the recorded video during the office-based survey and can get fatigue easily. An automated process can release the inspector's work load and ensure the consistent quality of the survey. However, a fully automated survey of varied structural discontinuities still remains as a challenge. This study aims to first identify the anomaly frames of the CCTV video, which contain the major anomalies captured from the internal surface of the pipe. Thus, the inspector can focus more on these anomaly frames. In this paper, an anomaly frame detection framework based on steerable pyramid autoencoder (SPAE) is proposed. The SPAE can generate discriminative representations to be used in the prediction. Both the parameter optimization and comparative studies for the proposed SPAE were carried out in this research. The experimental results demonstrate that this novel SPAE algorithm can achieve 0.984 accuracy and 0.984 F1-score, which outperforms other state-of-the-art methods selected for comparison. Thus, the proposed framework can significantly improve the accuracy and efficiency for anomaly frame detection, which will highly facilitate the pipeline condition assessment through the CCTV inspection. [Display omitted] • The paper proposes a steerable pyramid autoencoder based framework. • Anomaly frame detection is conducted on closed-circuit television videos. • The framework significantly facilitates the water pipeline condition assessment. • Steerable pyramid is integrated into the autoencoder based network. • The method develops great capability of representation learning and extraction. [ABSTRACT FROM AUTHOR]

To address the problem of excessive computation in the remote sensing scene classification (RSSC) algorithms, this paper proposes an efficient RSSC model, which is named BiShuffleNeXt. Firstly, we use the sandglass bottleneck to design the ShuffleNeXt as a context path for deep semantic information extraction. Secondly, the spatial path consists of three sandglass bottlenecks, each with a stride of 2, to retain more spatial information and obtain a high-resolution feature map. Finally, the feature combination module is designed to perform the fusion of bi-path features. Experimental results show that the proposed model achieves a classification accuracy of 95.93% on the NWPU-45 dataset, which is a 3.6% improvement over using the context path alone. The proposed method is tested on three datasets and outperforms ShuffleNet v2, MobileNetv2, Mobile-Former and ResNet-50 respectively. The floating point operations (FLOPs) of BiShuffleNeXt are only 298.25M, which greatly reduces the calculation amount while ensuring classification accuracy. [Display omitted] • A deep learning model named BiShuffleNeXt is proposed for efficient remote sensing scene classification. • Based on lightweight neural networks to ensure real-time inference of the model. • The proposed model separates the extraction of spatial and context information and deep semantic information of remote sensing images into two paths. [ABSTRACT FROM AUTHOR]

• A novel feature-aware network (FaNet) is proposed, which is applied to few-shot classification for strip steel surface defects. • Two feature-aware modules can effectively extract and enhance the high-dimensional feature information from a few image data. • The first publicly available few-shot classification dataset with large-scale strip steel surface defects is constructed (FSC-20). Accurate classification of surface defects is one of the most important factors in achieving quality inspection for strip steel. Most existing methods are based on fully-supervised learning, which requires a large number of labeled training data samples. In the manufacturing process, collecting defective samples is time-consuming and laborious. So it is very difficult to train a fully supervised model based on few labeled samples. In this paper, we propose a feature-aware network (FaNet) for a few shot defect classification, which can effectively distinguish new classes with a small number of labeled samples. In our proposed FaNet, we use ResNet12 as our baseline. The feature-attention convolution module (FAC) is applied to extract the comprehensive feature information from the base classes, as well as to fuse semantic information by capturing the long-range feature relationships between the upper and lower layers. Meanwhile, during the test phase, an online feature-enhance integration module (FEI) is adopted to average the noise from the support set and query set defect images, further enhancing image features among the different tasks. In addition, we construct a large-scale strip steel surface defects few shot classification dataset (FSC-20) with 20 different types. Experimental results show that the proposed method achieves the best performance compared to state-of-the-art methods for the 5-way 1-shot and 5-way 5-shot tasks. The dataset and code are available at: https://github.com/VDT-2048/FSC-20. [ABSTRACT FROM AUTHOR]

• A classification detection method based on rail defect features is proposed in this paper. • Based on extracted features, various defect detection models are built respectively. • Adaptive adjustment models are proposed for parameter thresholds. • A perceptron classification and learning model is designed to simply the detection processing. • The average detection time is 0.2 s/frame, and the average detection accuracy is higher than 97%. Many internal defects maybe arise in the rail of train, which will affect the safe driving of the high-speed railway. Considering the problem of intelligent detection of internal defects in railway, a classification detection method based on rail defect features is proposed in this paper. Firstly, rail defects features are extracted and classified based on their distribution features and contour morphological features. Then, rail defect detection models are built according to their defect features, and defects detection methods are designed. Finally, to improve detection parameters adaptability involved in the detection method, a threshold adjustment method is proposed based on their data distribution law. The proposed method can adjust the detection thresholds according to the design law to improve the detection accuracy. In addition, a data pre-screening perceptron classification and learning model is proposed for defect features of end face, weld reinforcement and lower crack of screw hole, which can transform the defects detection problem into a contour classification problem, and the detection performance can be improved by learning sample images. The rail data provided by the cooperation department are tested. Experiment results show that the detection accuracy rate of the proposed method is 97.3%, and the average detection time is 0.2 s/frame. The classifier experiment results also indicate that the proposed classifier shows good performance in recall and precision, which can be used to train and learn samples in defects detection processing. [ABSTRACT FROM AUTHOR]

In order to optimize a patient's cancer treatment, the prediction of the tumor's response to a planned radiotherapy is vital. For patients with rectal cancer, the quality of life depends heavily on the surgery leaving some patients with a stoma or other problems. If we would be able to predict the response upfront, we would be able to personalize the treatment and therefore preserve the QOL as much as possible. The Dworak tumor regression grade is a typical diagnostic tool to assess the tumor response of colorectal cancer patients. However, the Dworak grade is determined by a pathologist by inspection of the tumor biopsy without a dedicated measurement instrument. The measurement of the Dworak grade in an automated way by using the pre-operative CT scans, is a challenge. In this paper, we propose a novel methodology to measure the Dworak grade based on a customized random forest. We created a new random forest based on the methods from Breiman and Ho. These methods are further enhanced by evolutionary computation. We give the results of both our new method and other classical classification methods and highlight that the choice of classifier and knowledge is crucial. We extracted 111 radiomic features from 141 patients with colorectal cancer, of which 97 bad and 44 good responders. The evaluation gave a 77. 507 % accuracy and the cross-validation used for validation, gave a 67. 081 % accuracy. Our results were the best compared with the some other classification methods showing the need of a new method. • Colorectal cancer prediction after radiotherapy. • Radiomic features from 3D medical images. • Performance and fine-tuning Random Forest. • Machine learning as a feature selection tool. [ABSTRACT FROM AUTHOR]

When to block the taphole is a concern in blast furnace ironmaking process. Currently, the blocking time of the taphole mainly depends on expertise, causing poor reliability of the blocking operation. Therefore, this paper proposes a new method to intelligently determine the blocking time using molten iron flow images. Firstly, a monitoring system is designed to obtain molten iron flow images at the taphole. Secondly, the blocking time monitoring problem is transformed into a classification problem of molten iron image to obtain tapping stages, and a novel image classification module, which further promotes the network's representational ability by merging three feature extraction mechanisms in parallel, is proposed to accurately classify the images and obtain the tapping stages. Finally, a categorical correction model and a blocking time calculation model are constructed to obtain an accurate blocking time. Experimental results show that the proposed method can accurately monitor the taphole's blocking time. • A new monitoring method is proposed to realize the intelligent monitoring of the taphole blocking time in ironmaking process. • The blocking time monitoring problem is transformed into a classification problem of tapping stages and a statistical prediction problem. • A novel deep image classification module is proposed to classify molten iron flow images and obtain tapping stages. • According to the tapping stages, a categorical correction model and a remaining blocking time calculation model are constructed to obtain the accurate taphole blocking time. [ABSTRACT FROM AUTHOR]

Dust deposition and pollution are relevant issues in indoor environments, especially concerning human health and conservation of things and works. In this framework, several tools have been proposed in the last years in order to analyze dust deposition and extract useful information for addressing the phenomenon. In this paper, a novel approach for dust analysis and classification is proposed, employing machine learning and fuzzy logic to set up a simple and actual tool. The proposed approach is tested and compared with other already introduced similar techniques, in order to evaluate its performance. [ABSTRACT FROM AUTHOR]

• A novel data augmentation strategy named the overlapping sampling augmentation (OSA) technique is designed to generate large amounts of arc magnet data. • The multi-head attention strategy is introduced into the arc magnet defect detection model for the first time to highlight features that play an important role in defect detection. • A comprehensive experimental scheme including single category samples, multi categories samples, small samples, and the coexistence of noise and insufficient data is designed to verify the generalization and robustness of the proposed method from multiple perspectives. The results show that the arc magnet internal defect classification method has good industrial applications prospects. • Construct an acoustic-based detection system regarding arc magnet internal defect. Arc magnets are the key components of various motor machinery, and their internal defects detection is extremely significant for maintaining system performance and ensuring operational safety. In this paper, an end-to-end improved convolutional neural network (CNN) model based on multi-head attention is presented, where features that play a more important role in defect detection could be efficiently highlighted. In addition, owing to the characteristics of strong parallel working ability in multi-head attention, the training process is greatly accelerated. Meanwhile, to meet the requirements of the model on the amount of data, a data augmentation method is designed accordingly. Then, the performance of the constructed framework is verified in different test scenarios. Experiment results demonstrate that the presented approach owns superior inspection performance based on relatively fewer model parameters compared to other existing methods, even under the small sample, intense noise, and the coexistence of noise and insufficient data. [ABSTRACT FROM AUTHOR]

Breast cancer is known as the most common invasive cancer type among women and automatic breast cancer detection systems are in demand. Thus, various machine learning and pattern recognition techniques have been proposed to detect breast cancer. One of these techniques is the Bayes classifier. Naïve Bayesian (NB) is known to be a simple classifier, which is based on the Bayes theorem. There have been so many applications used in literature. In this paper, a new NB (weighted NB) classifier was proposed and its application on breast cancer detection was presented. Several experiments were conducted to evaluate the performance of the weighted NB on the breast cancer database. The experiments were realized with 5-fold cross validation test. Moreover, various performance evaluation techniques namely sensitivity, specificity and accuracy are considered. According to the experiments, the weighted NB obtained the following evaluation values. The calculated sensitivity, specificity and the accuracy values are 99.11%, 98.25%, and 98.54% respectively. Moreover, a comparison with the existing methods in the literature was presented. As a result, the performance of weighted NB is better than regular NB and many other existing methods. [ABSTRACT FROM AUTHOR]

Both fuzzy logic, as the basis of many inference systems, and Neural Networks, as a powerful computational model for classification and estimation, have been used in many application fields since their birth. These two techniques are somewhat supplementary to each other in a way that what one is lacking of the other can provide. This led to the creation of Neuro-Fuzzy systems which utilize fuzzy logic to construct a complex model by extending the capabilities of Artificial Neural Networks. Generally speaking all type of systems that integrate these two techniques can be called Neuro-Fuzzy systems. Key feature of these systems is that they use input–output patterns to adjust the fuzzy sets and rules inside the model. The paper reviews the principles of a Neuro-Fuzzy system and the key methods presented in this field, furthermore provides survey on their applications for technical diagnostics and measurement. [ABSTRACT FROM AUTHOR]

In this paper, a solid waste bin detection and waste level classification system that is rotation invariant is presented. First, possible locations and orientations of the bin are detected using Hough line detection. Then cross correlation is calculated to differentiate the true bin position and orientation from those of other similar objects. Next, features are extracted from the inside of the bin area and together with detected bin corners they are used to determine the bin’s waste level. A few features are also obtained from the outside of the bin area to check whether there is rubbish littered outside the bin. The proposed system was tested on shifted, rotated and unrotated bin images containing different level of waste. In the experiment, bin detection was treated separately from waste level classification. For bin detection, if 95% of the opening area is captured then the bin is considered detected correctly. The waste level classification is only considered for the correctly detected bins where the waste level is classified as empty, partially full or full. The system also checks the presence of rubbish outside the bin. In training, only images containing unrotated bin were used while in testing images containing both unrotated and rotated bin were used in equal number. The system achieves an average bin detection rate of 97.5% and waste level classification rate of 99.4% despite variations in the bin’s location, rotation and content. It is also robust against occlusion of the bin opening by large objects and confusion from square objects littered outside the bin. Its low average execution time suggests that the proposed method is suitable for real-time implementation. [ABSTRACT FROM AUTHOR]

This paper proposed an effort to investigate the suitability of input features and classifier for identifying thermal faults within electrical installations. The features are extracted from the thermal images of electrical equipment and classified using a multilayered perceptron (MLP) artificial neural network and support vector machine (SVM). In the experiments, the classification performances from various input features are evaluated. The commonly used classification performance indices, including sensitivity, specificity, accuracy, area under curve (AUC), receiver operating characteristic (ROC) and F -score are employed to identify the most suitable input feature as well as the best configuration of classifiers. The experimental results demonstrate that the combination of features set T max , T delta and DT bg produce the best input feature for thermal fault detection. In addition, the implementation of SVM using radial basis kernel function (RBF) produces slightly better performance than the MLP artificial neural network. [ABSTRACT FROM AUTHOR]

• A novel hybrid-based model is proposed for imbalanced ECG data classification. • The imbalanced classification task is solved in an iterative manner. • Samples in each class with high uncertainty are selected to form the training subset. • The label of the test sample is obtained by voting on predictions of all iterations. • The proposed model has excellent and stable performance on imbalanced datasets. This paper classifies non-ectopic (N), supraventricular ectopic (S), ventricular ectopic (V), and fusion (F) beats in the MIT-BIH arrhythmia database. The classification encounters serious class imbalance since the number of beats in N (majority class) with sample number above the average per class is heavily outnumbered than that in S, V, and F (minority classes) with sample number below the average per class. To address the class imbalance, a novel model based on active training subset selection and modified broad learning system (MBLS) is proposed. In each iteration, the MBLS trained with the current training subset is used to predict the class label of the test sample and actively select a new training subset for the next iteration. Finally, the class of the test sample is determined by voting on the predictions of all iterations. The experimental results show that our method has excellent performance and outperforms the existing methods. [ABSTRACT FROM AUTHOR]

This article deals with classification problems involving unequal probabilities in each class and discusses metrics to systems that use multilayer perceptrons neural networks (MLP) for the task of classifying new patterns. In addition we propose three new pruning methods that were compared to other seven existing methods in the literature for MLP networks. All pruning algorithms presented in this paper have been modified by the authors to do pruning of neurons, in order to produce fully connected MLP networks but being small in its intermediary layer. Experiments were carried out involving the E. coli unbalanced classification problem and ten pruning methods. The proposed methods had obtained good results, actually, better results than another pruning methods previously defined at the MLP neural network area. [ABSTRACT FROM AUTHOR]

Wheat product quality is closely related to wheat seed purity. Purity is an important factor that has a considerable impact on wheat product prices in grain storage silos. The aim of this paper was to introduce a machine vision based approach as a primarily step for fabricating an automatic wheat purity determination and grading device. Experimental data consists of 52 color, morphology, and texture characteristic parameters, extracted from images of samples, including four local wheat grades and eight common weed seeds growing in wheat fields of Iran, were used to build the classification models. A new algorithm that combines Imperialist Competitive Algorithm (ICA) and Artificial Neural Networks (ANNs) has been used for two purposes: to find the best characteristic parameters set and to create robust classification models. Based upon the results obtained from this study, the total classification rate of ICA-ANN approach for wheat grains vs. non-wheat seeds, wheat grain classes, and non-wheat seed classes was 96.25%, 87.50%, and 77.22%, respectively. [ABSTRACT FROM AUTHOR]

With the rapid expansion of e-commerce over the decades, the growth of the user generated content in the form of reviews is enormous on the Web. A need to organize the e-commerce reviews arises to help users and organizations in making an informed decision about the products. Opinion mining systems based on machine learning approaches are used online to categorize the customer opinion into positive or negative reviews. Different from previous approaches that employed single rule based or statistical techniques, we propose a hybrid machine learning approach built under the framework of combination (ensemble) of classifiers with principal component analysis (PCA) as a feature reduction technique. This paper introduces two hybrid models, i.e. PCA with bagging and PCA with Bayesian boosting models for feature based opinion classification of product reviews. The results are compared with two individual classifier models based on statistical learning i.e. logistic regression (LR) and support vector machine (SVM). We found that hybrid methods do better in terms of four quality measures like misclassification rate, correctness, completeness and effectiveness in classifying the opinion into positive and negative. [ABSTRACT FROM AUTHOR]

Highlights: [·] In some real applications, the status of some measures is not clearly known. [·] These measures are referred to as flexible measures. [·] This paper proposes a flexible SBM of efficiency in presence of flexible measures. [·] Two real cases show applicability of the proposed approach. [ABSTRACT FROM AUTHOR]

Microscopic analysis of blood-cells is an essential and vital task for the early diagnosis of life-threatening hematological disorders like blood cancer (leukemia). We have presented an effective and computationally efficient approach for automatically detecting and classifying Acute Lymphoblastic Leukemia (ALL) and Acute Myeloid Leukemia (AML). Currently, transfer learning has succeeded as a preferred approach in medical image analysis since it achieves excellent performance in a small database. This paper proposes a lightweight transfer-learning-based feature extraction followed by Support Vector Machine (SVM)-based classification technique for efficient ALL and AML detection. It yields a faster and more efficient system due to the depth-wise separable convolution, tunable multiplier, and inverted residual bottleneck structure. Moreover, the SVM-based classification improves the overall performance by optimizing the hyperplane location. Furthermore, the experimental results signify that our proposed system gains superior performance than others in all these three publicly available standard ALLIDB1, ALLIDB2, and ASH databases. • A computationally efficient blood cancer detection system is proposed. • Lightweight deep learning-based feature extraction is suggested to detect Leukemia. • Depth-wise separable convolution and inverted residual bottleneck makes it faster. • SVM-based classification boosts the performance by optimizing hyperplane location. • Experimental results signify the superiority of proposed deep learning model. [ABSTRACT FROM AUTHOR]

• Add a differential sequence to reduce the impact of noise in ECG signal. • Add a variable nonlinear layer to enhance the non-linear capability of the model. • Adaptive Learning rate is used to converge the training process. • The performance of proposed CNN is better than the other widely used algorithms. • The proposed algorithm can serve as an adjunct tool to assist clinicians in confirming their diagnosis. In this paper, an improved convolutional neural network (CNN) named Feature Selection CNN (FS-CNN) and Synthetic Minority Over-sampling Technique (SMOTE) named Multiply Stochastic SMOTE (MS-SMOTE) are proposed for imbalanced ECG multi-factor classification. FS-CNN includes convolutional expression layer, multi-residual block, and decision layer. The primary purpose of FS-CNN is to find the most influential features during automatic model optimization dynamically. MS-SMOTE combines the advantage of SMOTE and borderline SMOTE to balance the number of different categories. By dynamically selecting SMOTE and borderline SMOTE based on the proportion of data, the synthetic data can have a better category property. Comparing to other wide-used algorithms, our method is better than most other algorithms in many indicators. And in simulation 2, our method can deal with a large number of ECG multi-factor data classifications. Also, two self-controlled experiments are designed to examine how different parameters affect the result of the above problem. [ABSTRACT FROM AUTHOR]

Abstract: Computed tomography images are widely used in the diagnosis of ischemic stroke because of its faster acquisition and compatibility with most life support devices. This paper presents a new approach to automated detection of ischemic stroke using segmentation, midline shift and image feature characteristics, which separate the ischemic stroke region from healthy tissues in computed tomography images. The proposed method consists of five stages namely, pre-processing, segmentation, tracing midline of the brain, extraction of texture features and classification. The application of the proposed method for early detection of ischemic stroke is demonstrated to improve efficiency and accuracy of clinical practice. The results are quantitatively evaluated by a human expert. The average overlap metric, average precision and average recall between the results obtained using the proposed approach and the ground truth are 0.98, 0.99 and 0.98, respectively. A classification with accuracy of 98%, 97%, 96% and 92% has been obtained by SVM, k-NN, ANN and decision tree. [Copyright &y& Elsevier]

Abstract: This paper presents a novel approach of diagnosing actual analog circuits using improved support vector machines classifier (SVC) and this method falls into the category of fault dictionary. The stimulus is exerted on the circuit under test (CUT), and then the output responses are collected. Preprocessing technique is used to compress these responses and get feature samples. The fault classifier is based on the conventional “one against rest” SVC, which is then used to train these feature samples. In order to reduce the test time, the label analysis method for this classifier is employed. However, this simple method will generate a refusal area, which is then resolved by the introduction of space distance discriminant analysis and an apparent diagnosis performance improvement is thus achieved. Two actual experiments, based on data acquisition card (DAC) and digital signal processor (DSP) system respectively are demonstrated to validate the proposed method. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents novel vehicle detection and classification method by measuring and processing magnetic signal based on single micro-electro-mechanical system (MEMS) magnetic sensor. When a vehicle moves over the ground, it generates a succession of impacts on the earth’s magnetic field, which can be detected by single magnetic sensor. The magnetic signal measured by the magnetic sensor is related to the moving direction and the type of the vehicle. Generally, the recognition rate using single sensor detector is not high. In order to improve the recognition rate, a novel feature extraction algorithm and a novel vehicle classification and recognition algorithm are presented. The concavity and convexity areas, and the angles of concave and convex parts of the waveform are extracted. An improved support vector machine (ISVM) classifier is developed to perform vehicle classification and recognition. The effectiveness of the proposed approach is verified by outdoor experiments. [ABSTRACT FROM AUTHOR]

Abstract: Classification is a useful tool in identifying fault patterns. Generally, a good classification implementation is closely related to the effectiveness of data used. The word “effectiveness” implies that the data should be clean and the features indicating fault patterns should be properly selected. Unfortunately, data cleaning is not often implemented in reported work of fault pattern classifications. In this paper, a data processing algorithm is developed to achieve the effectiveness, which includes data cleaning followed by feature selection. A data cleaning algorithm is developed based on support vector machine and random sub-sampling validation. Candidate outliers are selected based on fraction values provided by the proposed data cleaning algorithm and final outliers are determined based on their removal impacts on classification performance. The feature selection algorithm adopts the classical sequential backward feature selection. The performance of the data cleaning algorithm is tested using three benchmark datasets. The tests show good capability of the data cleaning algorithm in identifying outliers for all datasets. The proposed data processing algorithm is adopted in the classification of the wear degree of pump impellers in a slurry pump system. The results show good effectiveness of sequentially using data cleaning and feature selection in addressing fault pattern classification problems. [Copyright &y& Elsevier]

• Sensitivity analysis of different transfer function components for different faults. • Correlation coefficient drawbacks investigation on fault detection. • Proposing a new index for the classification of different faults. • Investigation of fault classification accuracy of transfer function components. • Detection and classification of fault type, location and extent of faults. Frequency response analysis is one of the best methods to diagnose mechanical defects in the power transformer winding. In this paper, sensitivity analysis of four components of the transfer function (TF), i.e., amplitude, phase, real and imaginary components of TF is investigated for different defects. The classification of type, location, and extent of different defects is carried out using support vector machine with cubic polynomial kernel. The accuracy of different indices and four different components of TF are investigated. Moreover, a new index based on fitting performance is defined. The results show that despite widespread use and popularity of amplitude of TF, the phase, imaginary, and real components of TF, have more sensitivity to the defects respectively and are more accurate for classification. Also, compared with the results of some high usage indices, the proposed index has the best classification accuracy. [ABSTRACT FROM AUTHOR]

• Early prediction for mammography. • To diagnose the breast cancer at the earliest to reduce the mortality rate. • This paper proposes an expert system for the diagnosis of breast cancer disease using DE-RBF-KELM method. • The proposed method is effectively classifying the abnormal classes of mammograms. Breast cancer disease is a major public health problem among women worldwide. This article proposes an expert system for the diagnosis of breast cancer disease based on an evolutionary algorithm known as Differential Evolution (DE) of a Radial-Based Function Kernel Extreme Learning Machines (RBF-KELM). In the structure of the RBF-KELM, there are two adjustable parameters of the RBF-kernel which are the penalty parameter C and the RBF-kernel's parameter (σ). These parameters play a major role in the efficiency of RBF-KELM. In this study, the optimal values of these parameters have been obtained using a differential evolution (DE) algorithm. To validate the effectiveness of the suggested approach, DE-RBF-KELM was examined on the two datasets: The Mammographic Image Analysis Society (MIAS) and the wisconsin breast cancer database (WBCD) and the results were satisfactory compared to conventional approaches. [ABSTRACT FROM AUTHOR]