Common Feature Extraction Methods Research Articles

Global optimal Ramanujan spectrum: A feature extraction method without pseudo-monotonicity

As a common state feature extraction method, spectral analysis methods have been widely used in various fields. However, due to the influence of fixed envelope, the existing spectral analysis method is not outstanding in extracting state features, and it is difficult to extract state features under strong interference. In this paper, a new feature extraction method, called global optimal Ramanujan spectrum (GORS), is proposed. GORS method innovatively uses iterative generalized envelope to improve the accurate enhancement of weak fault features and avoid the defect of unstable feature extraction performance caused by fixed envelope. Meanwhile, an evaluation index without pseudo-monotonicity, logarithmic Ramanujan spectral signal-to-noise ratio (lg(RSSNR)), is defined for the first time, which is used to help select the optimal envelope results and adaptively determine the optimal iterative envelope number, so as to realize the global optimization of spectral analysis results and maximize the ability of state feature extraction. (Pseudo-monotonicity of feature index (PMFI) is defined as: there is no stable correspondence between the change law of characteristic index and the information content of state characteristics). The evaluation results of fault characteristic strength (FCS) by different methods show that GORS method has good anti-interference ability and is an effective state feature extraction method.

SVM-based Support Vector Type Recognition Machine for Smart Things in Soccer Training Motion Recognition

With the development of IoT technology, machine learning, and other artificial intelligence technologies, there have been many related technologies applied to the sports industry. Soccer, as the world’s number one sport, has a wide range of popularity, a high degree of attention, and a high degree of commercialization. In the traditional soccer training action recognition methods, there is insufficient collection and in-depth profiling of real data, and what is not available is soccer movement action capture and recognition based on kinematic knowledge. To address the above shortcomings, this study designs a Support Vector Machine SVM-based intelligent IoT-type soccer training movement recognition and evaluation framework, and constructs a machine learning algorithmic model to recognize, evaluate and analyze soccer training movements. Common feature extraction methods are suitable for recognizing most monotonous movements, but soccer movements are highly variable and athletes’ ankle movements are flexible and changeable. The actual acquired soccer data streams are noisy and the data patterns are not obvious, and the performance of the model to recognize the data will be degraded. To extract the effective feature values in the complex data stream and improve the correct degree of pattern recognition, the classification pattern of attitude angle type solving + SVM classification algorithm is constructed. The experimental results show that the designed algorithmic pattern based on the posture angular pattern solving + SVM classification algorithm pattern for soccer training movement recognition can reach 90% accuracy in recognizing different movements, which is extremely suitable for the recognition of soccer training movements.

Research on Hail Mechanism Features Based on Dual-Polarization Radar Data

Hail is a type of severe convective weather disaster characterized by abundant water vapor and strong updrafts, resulting in intense and high reflectivity echoes in hail clouds, often accompanied by an overhanging form. Although hail research has made great progress, it is still challenging to achieve accurate identification of hail. Compared with traditional radar, dual-polarization radar can output a variety of polarization parameters and provide information about the shape and phase of precipitation particles, which is conducive to the identification of hail particles. In this study, dual-polarization radar data are used to explore more hail features from various perspectives, starting with the morphological characteristics of hail clouds and using common feature extraction methods in the field of image processing. A comprehensive approach to high-dimensional features is developed. Using machine learning methods, hail identification models are constructed in both the traditional mechanism feature space and the new feature space constructed in this study. Experimental results strongly confirm the significant effectiveness of the five-dimensional new mechanism features developed in this paper for hail identification.

Cloud ensemble learning for fault diagnosis of rolling bearings with stochastic configuration networks

The scarcity of fault samples poses a significant challenge for fault diagnosis of rolling bearings in industrial environment. Conventional fault diagnosis methods struggle to achieve satisfactory results in the few-shot scenarios. Furthermore, common entropy-based feature extraction methods fail to adaptively represent the uncertainty information hidden in vibration signals. To overcome these issues, this article proposes a stochastic configuration network-based cloud ensemble learning (SCN-CEL). Firstly, a cloud feature extraction method is developed to effectively capture fault information from vibration signals while accounting for their inherent uncertainty based on backward cloud generator (BCG) of cloud model (CM), without requiring hyperparameter settings. Subsequently, a cloud oversampling (COS) method is proposed to augment the feature space of limited samples and generate sufficient samples for improving diagnostic accuracy based on bidirectional cloud generator. Finally, we introduce an ensemble model that combines SCNs with multiple constrained COS to comprehensively characterize uncertain fault information and advance the generalization of diagnosis machine. By harnessing the constructive incremental learning of SCN, SCN-CEL guarantees both efficient modeling and accurate prediction for bearing fault diagnosis. Extensive experiments evaluate the effectiveness of each module in SCN-CEL and demonstrate its favorable performance in distinguishing fault categories of rolling bearings in the few-shot scenarios.

A fine recognition method of strawberry ripeness combining Mask R-CNN and region segmentation.

As a fruit with high economic value, strawberry has a short ripeness period, and harvesting at an incorrect time will seriously affect the quality of strawberries, thereby reducing economic benefits. Therefore, the timing of its harvesting is very demanding. A fine ripeness recognition can provide more accurate crop information, and guide strawberry harvest management more timely and effectively. This study proposes a fine recognition method for field strawberry ripeness that combines deep learning and image processing. The method is divided into three stages: In the first stage, self-calibrated convolutions are added to the Mask R-CNN backbone network to improve the model performance, and then the model is used to extract the strawberry target in the image. In the second stage, the strawberry target is divided into four sub-regions by region segmentation method, and the color feature values of B, G, L, a and S channels are extracted for each sub-region. In the third stage, the strawberry ripeness is classified according to the color feature values and the results are visualized. Experimental results show that with the incorporation of self-calibrated convolutions into the Mask R-CNN, the model's performance has been substantially enhanced, leading to increased robustness against diverse occlusion interferences. As a result, the final average precision (AP) has improved to 0.937, representing a significant increase of 0.039 compared to the previous version. The strawberry ripeness classification effect is the best on the SVM classifier, and the accuracy under the combined channel BGLaS reaches 0.866. The classification results are better than common manual feature extraction methods and AlexNet, ResNet18 models. In order to clarify the role of the region segmentation method, the contribution of different sub-regions to each ripeness is also explored. The comprehensive results demonstrate that the proposed method enables the evaluation of six distinct ripeness levels of strawberries in the complex field environment. This method can provide accurate decision support for strawberry refined planting management.

State of Charge Estimation for Batteries Based on Common Feature Extraction and Transfer Learning

The state of charge (SOC) of a battery is a key parameter of electrical vehicles (EVs). However, limited by the lack of computing resources, the SOC estimation strategy used in vehicle-mounted battery management systems (V-BMS) is usually simplified. With the development of the new energy vehicle big data platforms, it is possible to obtain the battery SOC through cloud-based BMS (C-BMS). In this paper, a battery SOC estimation method based on common feature extraction and transfer learning is proposed for C-BMS applications. Considering the diversity of driving cycles, a common feature extraction method combining empirical mode decomposition (EMD) and a compensation strategy for C-BMS is designed. The selected features are treated as the new inputs of the SOC estimation model to improve the generalization ability. Subsequently, a long short-term memory (LSTM) recurrent neural network is used to construct a basic model for battery SOC estimation. A parameter-based transfer learning method and an adaptive weighting strategy are used to obtain the C-BMS battery SOC estimation model. Finally, the SOC estimation method is validated on laboratory datasets and cloud platform datasets. The maximum root-mean-square error (RMSE) of battery SOC estimation with the laboratory dataset is 2.2%. The maximum RMSE of battery pack SOC estimation on two different electric vehicles is 1.3%.

Classification of intracranial hemorrhage CT images based on texture analysis using ensemble-based machine learning algorithms: A comparative study

Blood vessels in the brain tissue can leak or burst, causing an intracranial hemorrhage, a potentially fatal disorder. One of the critical steps in brain stroke imaging is the categorization of intracranial hemorrhages. This article aims to perform texture-based intracranial hemorrhage computed tomography image classification with ensemble and machine learning classifiers using local binary pattern, local ternary pattern, and Weber local descriptor approach. The most common texture feature extraction method is a local binary pattern; however, other methods such as local ternary pattern, Weber local descriptor, and local binary pattern are also utilized for texture feature extraction. For all the extracted texture codes, histograms are applied, and finally, single feature vector texture codes by performing the concatenation of the bin. The efficacy of feature descriptors for classifying images into abnormal and normal classes was assessed using ten distinct classifiers. The result shows that Weber's local descriptor shows promising results in texture code classification for ensemble-based classification. For all texture-based methods, classification accuracy was better than standard machine learning. According to the findings of the experiment, the Random Forest classifier outperformed all other classifiers in terms of classification accuracy (86.55%), recall (86.31%), precision (87.23%), sensitivity (86.31%), specificity (86.81%), and F1-score (86.77%) for the Weber local descriptor.

State of the art on vibration signal processing towards data‐driven gear fault diagnosis

Gear fault diagnosis (GFD) based on vibration signals is a popular research topic in industry and academia. This paper provides a comprehensive summary and systematic review of vibration signal-based GFD methods in recent years, thereby providing insights for relevant researchers. The authors first introduce the common gear faults and their vibration signal characteristics. The authors overview and compare the common feature extraction methods, such as adaptive mode decomposition, deconvolution, mathematical morphological filtering, and entropy. For each method, this paper introduces its idea, analyses its advantages and disadvantages, and reviews its application in GFD. Then the authors present machine learning-based methods for gear fault recognition and emphasise deep learning-based methods. Moreover, the authors compare different fault recognition methods. Finally, the authors discuss the challenges and opportunities towards data-driven GFD.

Developing Tensor-Based Common and Special Feature Analysis for Comprehensive Monitoring of Complex Batch Processes

To overcome the limitation of unfolding-based methods and handle the multiple data set and limited data problems in the complex processes, such as multigrade batch processes, a novel tensor-based common and special feature extraction method and a comprehensive monitoring framework are proposed. In the proposed method, the uneven-length three-dimensional data are directly analyzed by the comprehensive tensor-based method without unfolding. To handle the multiple data set modeling problem, the tensor-based common feature extraction methods are first proposed to obtain the common features shared among different grades. The special features are sequentially determined by conducting tensor principal component analysis (PCA) on the residuals of each grade. The data are thus divided into common, special, and residual subspaces. Three monitoring statistics are established respectively in each subspace for online fault detection. The merits and effectiveness of the proposed method are demonstrated by an injection molding process with both even-length and uneven-length data in comparison with traditional methods.

ScIAE: an integrative autoencoder-based ensemble classification framework for single-cell RNA-seq data.

Single-cell RNA sequencing (scRNA-seq) allows quantitative analysis of gene expression at the level of single cells, beneficial to study cell heterogeneity. The recognition of cell types facilitates the construction of cell atlas in complex tissues or organisms, which is the basis of almost all downstream scRNA-seq data analyses. Using disease-related scRNA-seq data to perform the prediction of disease status can facilitate the specific diagnosis and personalized treatment of disease. Since single-cell gene expression data are high-dimensional and sparse with dropouts, we propose scIAE, an integrative autoencoder-based ensemble classification framework, to firstly perform multiple random projections and apply integrative and devisable autoencoders (integrating stacked, denoising and sparse autoencoders) to obtain compressed representations. Then base classifiers are built on the lower-dimensional representations and the predictions from all base models are integrated. The comparison of scIAE and common feature extraction methods shows that scIAE is effective and robust, independent of the choice of dimension, which is beneficial to subsequent cell classification. By testing scIAE on different types of data and comparing it with existing general and single-cell-specific classification methods, it is proven that scIAE has a great classification power in cell type annotation intradataset, across batches, across platforms and across species, and also disease status prediction. The architecture of scIAE is flexible and devisable, and it is available at https://github.com/JGuan-lab/scIAE.

Neural Decoding of EEG Signals with Machine Learning: A Systematic Review.

Electroencephalography (EEG) is a non-invasive technique used to record the brain’s evoked and induced electrical activity from the scalp. Artificial intelligence, particularly machine learning (ML) and deep learning (DL) algorithms, are increasingly being applied to EEG data for pattern analysis, group membership classification, and brain-computer interface purposes. This study aimed to systematically review recent advances in ML and DL supervised models for decoding and classifying EEG signals. Moreover, this article provides a comprehensive review of the state-of-the-art techniques used for EEG signal preprocessing and feature extraction. To this end, several academic databases were searched to explore relevant studies from the year 2000 to the present. Our results showed that the application of ML and DL in both mental workload and motor imagery tasks has received substantial attention in recent years. A total of 75% of DL studies applied convolutional neural networks with various learning algorithms, and 36% of ML studies achieved competitive accuracy by using a support vector machine algorithm. Wavelet transform was found to be the most common feature extraction method used for all types of tasks. We further examined the specific feature extraction methods and end classifier recommendations discovered in this systematic review.

Pre-trained CNNs Models for Content based Image Retrieval

Content based image retrieval (CBIR) systems is a ‎common recent method for image retrieval and is ‎based mainly ‎on two pillars extracted features and similarity measures. Low ‎level image presentations, ‎based on colour, texture and shape ‎properties are the most common feature extraction methods used ‎by ‎traditional CBIR systems. Since these traditional handcrafted ‎features require good prior domain ‎knowledge, inaccurate ‎features used for this type of CBIR systems may widen the ‎semantic gap and ‎could lead to very poor performance retrieval ‎results. Hence, features extraction methods, which ‎are ‎independent of domain knowledge and have automatic ‎learning capabilities from input image are ‎highly useful. Recently, ‎pre-trained deep convolution neural networks (CNN) with ‎transfer learning ‎facilities have ability to generate and extract ‎accurate and expressive features from image data. Unlike ‎other ‎types of deep CNN models which require huge amount of data ‎and massive processing time ‎for training purposes, the pre-‎trained CNN models have already trained for thousands of ‎classes of large-scale data, including huge ‎images and their ‎information could be easily used and transferred. ResNet18 ‎and ‎SqueezeNet are successful and effective examples of pre-‎trained CNN models used recently in many ‎machine learning ‎applications, such as classification, clustering and object ‎recognition. In this ‎study, we have developed CBIR systems ‎based on features extracted using ResNet18 and SqueezeNet ‎pre-‎trained CNN models. Here, we have utilized these pre-trained ‎CNN models to extract two groups of features ‎that are stored ‎separately and then later are used for online image searching and ‎retrieval. Experimental ‎results on two popular image datasets ‎Core-1K and GHIM-10K show that ResNet18 features ‎based on ‎the CBIR method have overall accuracy of 95.5% and 93.9% for ‎the two datasets, respectively, which ‎greatly outperformed the ‎traditional handcraft features based on the CBIR method.‎

Optimal channel-based sparse time-frequency blocks common spatial pattern feature extraction method for motor imagery classification.

Common spatial pattern (CSP) as a spatial filtering method has been most widely applied to electroencephalogram (EEG) feature extraction to classify motor imagery (MI) in brain-computer interface (BCI) applications. The effectiveness of CSP is determined by the quality of interception in a specific time window and frequency band. Although numerous algorithms have been designed to optimize CSP by splitting the EEG data with a sliding time window and dividing the frequency bands with a set of band-pass filters, simultaneously. However, they did not consider the drawbacks of the rapid increase in data volume and feature dimensions brought about by this method, which would reduce the classification accuracy and calculation efficiency of the model. Therefore, we propose an optimal channel-based sparse time-frequency blocks common spatial pattern (OCSB-CSP) feature extraction method to improve the model classification accuracy and computational efficiency. Comparative experiments on two public EEG datasets show that the proposed method can quickly select significant time-frequency blocks and improve classification performance. The average classification accuracies are higher than those of other winners' methods, providing a new idea for the improvement of BCI applications.

Identification method of cashmere and wool based on texture features of GLCM and Gabor

The common texture feature extraction method is only in spatial or frequency domain, leading to insufficient texture information and low accuracy. The main aim of this paper is to present a novel texture feature analysis method based on gray level co-occurrence matrix and Gabor wavelet transform to sufficiently extract texture feature of cashmere and wool fibers. Firstly, the gray level co-occurrence matrix is constructed to calculate the four texture feature vectors including of contrast, angular second moment, dissimilarity and energy in spatial domain, and four texture feature vectors, which are contrast, angular second moment, mean and entropy, in frequency domain is obtained through Gabor wavelet transform and Gray-Scale difference statistics method. Then, because the contrast and angle second moment are used as descriptors of fiber image in both spatial and frequency domain, they are fused respectively by introducing a weight to make linear addition, making eight feature values compose a 6-dimensional feature vector. Finally, these feature vectors are fed into the Fisher classifier. The experimental results show that the identification accuracy of the proposed algorithm is improved by 0.682% compared to use 8-dimensional feature vectors describing the sample image. It verifies that the fused method based on texture feature in spatial and frequency domain is an effective approach to identify fibers of cashmere and wool.

Saliency Maps-Based Convolutional Neural Networks for Facial Expression Recognition

Facial expression recognition (FER) is one of the important research contents in affective computing. It plays a key role in many application fields of human life. As a most common expression feature extraction method, the convolutional neural network (CNN) has the following main limitation. Due to the fact that the CNN network lacks the visual attention guidance, when it gets expression information it brings background noises, resulting in the lower recognition accuracy. In order to simulate the attention mechanism in human visual system, a salient feature extraction model is proposed, including the dilated inception module, the Difference of Gaussian (DOG) module, and the multi-indicator saliency prediction module. This model can effectively reflect the key facial information through the increase of the receptive field, the acquisition of multiscale features, and the simulation of human vision. In addition, a novel FER method for one single person is proposed. With the prior knowledge of saliency maps and the multilayer deep features in the CNN network, the recognition accuracy is improved by obtaining more targeted and more complete deep expression information. The experimental results of saliency prediction, action unit (AU) detection, and smile intensity estimation on the CAT2000, the CK+, and the BP4D databases prove that the proposed method improves the FER performance and is more effective than the existing approaches.

WLnet: Towards an Approach for Robust Workload Estimation Based on Shallow Neural Networks

Electroencephalography (EEG) is a non-invasive technology used for the human brain-computer interface. One of its important applications is the evaluation of the mental state of an individual, such as workload estimation. In previous works, common spatial pattern feature extraction methods have been proposed for the EEG-based workload detection. Recently, several novel methods were introduced to detect EEG pattern workloads. However, it is still unknown which one of these methods is the one that offers the best performance for the workload EEG pattern feature detection. In this article, four methods were used to extract workload EEG features: (a) common spatial pattern feature extraction; (b) temporally constrained sparse group spatial pattern feature extraction; (c) EEGnet; and (d) the new proposed shallow convolutional neural network for workload estimation (WLnet). The classification accuracy of these four methods was compared. Experimental results demonstrate that the proposed WLnet achieved the best detection accuracy in both stress and non-stress conditions. We believe that the proposed methods may be relevant to real-life applications of mental workload estimation.

Fault Diagnosis of an Analog Circuit Based on Hierarchical DVS

Analog circuit fault diagnosis technology is widely used in the diagnosis of various electronic devices. The basic strategy is to extract circuit fault characteristics and then to use a clustering algorithm for diagnosis. The discrete Volterra series (DVS) is a common feature extraction method; however, it is difficult to calculate its parameters. To solve the problem of feature extraction in fault diagnosis, we propose an improved hierarchical Levenberg–Marquardt (LM)–DVS algorithm (IDVS). First, the DVS is simplified on the basis of the hierarchical symmetry of the memory parameters, the LM strategy is used to optimize the coefficients, and a Bayesian information criterion based on the symmetry of entropy is introduced for order selection. Finally, we propose a fault diagnosis method by combining the improved DVS algorithm and a condensed nearest neighbor algorithm (CNN) (i.e., the IDVS–CNN method). A simulation experiment was conducted to verify the feature extraction and fault diagnosis ability of the IDVS–CNN. The results show that the proposed method outperforms conventional methods in terms of the macro and micro F1 scores (0.903 and 0.894, respectively), which is conducive to the efficient application of fault diagnosis. In conclusion, the improved method in this study is helpful to simplify the calculation of the DVS parameters of circuit faults in analog electronic systems, and provides new insights for the prospective application of circuit fault diagnosis, system modeling, and pattern recognition.

Accelerometer-Based Key Generation and Distribution Method for Wearable IoT Devices

With the fast development of wearable IoT devices, their applications are becoming more and more pervasive, ranging from social networking, payment, and navigation to health and activity monitoring. The security of the communication between these devices is essential to protect the transmitted sensitive information from tampering and eavesdropping. With the integration of accelerometers into wearable IoT devices, the gait-based biometric cryptography technology has emerged as a data securing tool for wearables. This article proposes a lightweight noise-based group key generation method, which utilizes the noise signals imposed on the raw acceleration signals to generate an M-bit key with high randomness and bit generation rate. Moreover, a signed sliding window coding (SSWC)-based common feature extraction method was designed to extract the common feature for sharing the generated M-bit key among devices worn on different body parts. Finally, a fuzzy vault-based group key distribution system was implemented and evaluated using a public data set. The performed comprehensive analysis of the proposed key generation and distribution method proved that the binary keys generated via the introduced noise-based procedure have high entropy and can pass both the NIST and Dieharder statistical tests with high efficiency. The experimental results obtained prove the robustness of the proposed SSWC-based common feature extraction method in terms of the similarity and discriminability of intra- and inter-class features, respectively.

Fault diagnosis based on deep learning by extracting inherent common feature of multi-source heterogeneous data

Fault diagnosis can provide a basic guarantee for safe operation of industrial equipment. Deep learning has attracted much attention from experts in the field of fault diagnosis because of its powerful feature representation ability. But traditional deep learning methods cannot well extract common feature from multi-source heterogeneous data which is the inherent character of the monitored object. Using only one kind of heterogeneous data for deep learning fault diagnosis will inevitably result in poor diagnosis accuracy. Aiming at this problem, this article proposes a deep common feature extraction method by designing a fusion network with alternating optimization mechanism. The rough features extracted independently from two kinds of heterogeneous data are used to train the designed fusion network in an alternative optimization way. A new loss function required by alternative optimization is established; thus, all the networks can be tuned globally. The deep common features of multi-source heterogeneous data can be well extracted by alternating optimization training process of the fusion network, which improves the accuracy of deep learning fault diagnosis method. Experiments for rolling bearings fault diagnosis testify the effectiveness of the proposed algorithm.

Multimode Process Monitoring and Fault Diagnosis Based on Tensor Decomposition

Nowadays, many industrial processes generate large amounts of multimode data, which generally have a natural tensor structure, causing some faults invisible with traditional process monitoring (PM) and fault diagnosis (FD) methods. Tensor decomposition (TD) is a more practical approach for its effectiveness in solving high dimensionality problems as well as indicating the links between different modes. This paper proposes a common and individual feature extraction method based on TD, which identifies and separates the common and individual features from multimode data. The newly proposed approach is applied to a typical multimode hot strip mill process (HSMP), where common and individual feature for all steel products are existing. The final results indicate that the proposed approach can accurately detect and identify different faults in the HSMP.