Search

In this paper, we propose an end-to-end approach to addressing QRS complex detection and measurement of Electrocardiograph (ECG) paper using convolutional neural networks (CNNs). Unlike conventional detection solutions that convert images to digital data, our method can directly detect QRS complex in images using Faster-RCNN, then the R-peak can be located and measured through a CNN. Validated by clinical ECG data in the St.-Petersburg Institute of Cardiological Technics 12-lead Arrhythmia Database and real ECG paper from Peking University People’s Hospital, the proposed method can achieve the recall of 98.32%, the precision of 99.01% in detecting and 0.012 mv of mean absolute error in measuring. Experimental results demonstrate the superior performance of our method over conventional solutions, which would pave the way to detect and measure ECG paper using CNNs.

To improve the accuracy of user implicit rating prediction, we combine the traditional latent factor model (LFM) and bidirectional gated recurrent unit neural network (BiGRU) model to propose a hybrid model that deeply mines the latent semantics in the unstructured content of the text and generates a more accurate rating matrix. First, we utilize the user’s historical behavior (favorites records) to build a user rating matrix and decompose the matrix to obtain the latent factor vectors of users and literature. We also apply the BERT model for word embedding of the research papers to obtain the sequence of word vectors. Then, we apply the BiGRU with the user attention mechanism to mine the research paper textual content and to generate the new literature latent feature vectors that are used to replace the original literature latent factor vectors decomposed from the rating matrix. Finally, a new rating matrix is generated to obtain users’ ratings of noninteractive research papers and to generate the recommendation list according to the user latent factor vector. We design experiments on the real datasets and verify that the research paper recommendation model is superior to traditional recommendation models in terms of precision, recall, F1-value, coverage, popularity and diversity.

Nowadays, there is a rapidly increasing number of conferences and journals in computer science that make a lot of challenges for researchers to find an appropriate venue to submit their scientific work. There is a need for a recommendation system that can support researchers for a better process of paper submission. In this paper, we present an efficient approach for building such a recommendation model by using embedding methods, Global Vector (GloVe) 1 created by Pennington et al. [1] and FastText 2 proposed by Facebook [2], Convolutional Neural Network (CNN) [3], and LSTM. The experimental results show that the combination of CNNs and FastText, CNN + FastText, can achieve the best performance in terms of the Top 1 Accuracy compared with other techniques, including the S2RSCS model, as presented in [4]. Moreover, the performance by using GloVe or FastText is much better, faster, and more stable than S2RSCS in most cases.

We present Siam R-CNN, a Siamese re-detection architecture which unleashes the full power of two-stage object detection approaches for visual object tracking. We combine this with a novel tracklet-based dynamic programming algorithm, which takes advantage of re-detections of both the first-frame template and previous-frame predictions, to model the full history of both the object to be tracked and potential distractor objects. This enables our approach to make better tracking decisions, as well as to re-detect tracked objects after long occlusion. Finally, we propose a novel hard example mining strategy to improve Siam R-CNN's robustness to similar looking objects. Siam R-CNN achieves the current best performance on ten tracking benchmarks, with especially strong results for long-term tracking. We make our code and models available at www.vision.rwth-aachen.de/page/siamrcnn., Comment: CVPR 2020 camera-ready version

The problem of classifying a research article into one or more fields of science is of particular importance for the academic search engines and digital libraries. A robust classification algorithm offers the users a wide variety of useful tools, such as the refinement of their search results, the browsing of articles by category, the recommendation of other similar articles, etc. In the current literature we encounter approaches which attempt to address this problem without taking into consideration important parameters such as the previous history of the authors and the categorization of the scientific journals which publish the articles. In addition, the existing works overlook the huge volume of the involved academic data. In this paper, we expand an existing effective algorithm for research articles classification, and we parallelize it on Apache Spark –a parallelization framework which is capable of sharing large amounts of data into the main memory of the nodes of a cluster– to enable the processing of large academic datasets. Furthermore, we present data manipulation methodologies which are useful not only for this particular problem, but also for most parallel machine learning approaches. In our experimental evaluation, we demonstrate that our proposed algorithm is considerably more accurate than the supervised learning approaches implemented within the machine learning library of Spark, whereas it outperforms them in terms of execution speed by a significant margin.

In this paper, a framework for human hand movement recognition is implemented for humans to play rock-paper-scissors game interactively against computer using a webcam. The game is the classic rock-paper-scissors game; however, first player is human and second player is computer. Human hand movements are recognized using a webcam, and the game is fully controlled using hand gestures. A dataset is constructed in this work by collecting hand images of different people. The dataset contains rock, paper, and scissors movements of different humanbeings. There are 8 photos for each movement and person using their left and right hands. The dataset is constructed using a controlled environment with black background and close up hand images. To recognize player movements two different feature extraction methods and two different classification methods are used and compared. Histogram of oriented gradients feature extraction method combined with support vector machines are found to be effective. It presents fast and effective computer vision framework for seamless playing experience with a sensitivity of %92.9.

Falsification of medicines is a big problem in many developing countries, where technological infrastructure is inadequate to detect these harmful products. We have developed a set of inexpensive paper cards, called Paper Analytical Devices (PADs), which can efficiently classify drugs based on their chemical composition, as a potential solution to the problem. These cards have different reagents embedded in them which produce a set of distinctive color descriptors upon reacting with the chemical compounds that constitute pharmaceutical dosage forms. If a falsified version of the medicine lacks the active ingredient or includes substitute fillers, the difference in color is perceivable by humans. However, reading the cards with accuracy takes training and practice, which may hamper their scaling and implementation in low resource settings. To deal with this, we have developed an automatic visual recognition system to read the results from the PAD images. At first, the optimal set of reagents was found by running singular value decomposition on the intensity values of the color tones in the card images. A dataset of cards embedded with these reagents is produced to generate the most distinctive results for a set of 26 different active pharmaceutical ingredients (APIs) and excipients. Then, we train two popular convolutional neural network (CNN) models, with the card images. We also extract some "hand-crafted" features from the images and train a nearest neighbor classifier and a non-linear support vector machine with them. On testing, higher-level features performed much better in accurately classifying the PAD images, with the CNN models reaching the highest average accuracy of over 94\%., Comment: in Proc. IEEE Winter Conference on Applications of Computer Vision (WACV), 2016

Monetary transactions are integral part of our day to day activities, so currency recognition has become one of the active research area at present and it has vast potential applications. In this paper we introduced a system to recognize and classify four different currencies using computer vision. As there are 200+different currencies used in different countries around the world. The technology of currency recognition aims to search and extract the visible as well as hidden marks on paper currency for efficient classification. The features are extracted based on color, texture and shape for four different currencies and they are classified using Artificial Neural network.

This paper proposes a new method for query expansion based on bidirectional extraction of phrases as word n-grams from research paper titles. The proposed method aims to extract information relevant to users’ needs and interests and thus to provide a useful system for technical paper retrieval. The outcome of proposed method are the trigrams as phrases that can be used for query expansion. First, word trigrams are extracted from research paper titles. Second, a co-occurrence graph of the extracted trigrams is constructed. To construct the co-occurrence graph, the direction of edges is considered in two ways: forward and reverse. In the forward and reverse co-occurrence graphs, the trigrams point to other trigrams appearing after and before them in a paper title,respectively. Third, Jaccard similarity is computed between trigrams as the weight of the graph edge. Fourth, the weighted version of PageRank is applied. Consequently, the following two types of phrases can be obtained as the trigrams associated with the higher PageRank scores. The trigrams of the one type, which are obtained from the forward co-occurrence graph, can form a more specific query when users add a technical word or words before them. Those of the other type, obtained from the reverse co-occurrence graph, can form a more specific query when users add a technical word or words after them. The extraction of phrases is evaluated as additional features in the paper title classification task using SVM. The experimental results show that the classification accuracy is improved than the accuracy achieved when the standard TF-IDF text features are only used. Moreover, the trigrams extracted by the proposed method can be utilized to expand query words in research paper retrieval.

Introduction: One attractive research area in the computer science field is soft biometrics. Aim: To Identify a person’s gender from an iris image when such identification is related to security surveillance systems and forensics applications. Methods: In this paper, a robust iris gender-identification method based on a deep convolutional neural network is introduced. The proposed architecture segments the iris from a background image using the graph-cut segmentation technique. The proposed model contains 16 subsequent layers; three are convolutional layers for feature extraction with different convolution window sizes, followed by three fully connected layers for classification. Results: The original dataset consists of 3,000 images, 1,500 images for men and 1,500 images for women. The augmentation techniques adopted in this research overcome the overfitting problem and make the proposed architecture more robust and immune from simply memorizing the training data. In addition, the augmentation process not only increased the number of dataset images to 9,000 images for the training phase, 3,000 images for the testing phase and 3,000 images for the verification phase but also led to a significant improvement in testing accuracy, where the proposed architecture achieved 98.88%. A comparison is presented in which the testing accuracy of the proposed approach was compared with the testing accuracy of other related works using the same dataset. Conclusion: The proposed architecture outperformed the other related works in terms of testing accuracy.

In this paper, we propose a new approach to zone identification which is based on considering features with high semantic richness. Out of the scenarios of selecting features for identifying a zone based on classifying the sentences in a text, we came to notice that in the scenario where specialized names belonging to a text’s domain and mode of the verbs together with reduced versions of conventional features, including history, are taken into account, an accuracy rate of 61% (resp. 81%) is obtained which is higher than that belonging to Liakata’s (resp. Fisas’s) approach. Also, to have a genuine comparison, both Liakata’s and Fisas’s corpora are used in our experiments. Such accuracy is obtained at the place where less computational cost for extracting the features was decreased. In order to improve the accuracy of zone identification, a decision-level fusion process based on combining the results of separate classifiers, was considered. With regard to this, two fusion techniques of ‘majority voting’ and ‘average of probabilities’ were used. Experimentations show the fact that ensemble of ‘Logistic Regression’, ‘Support Vector Machine’ and ‘Neural Network’ as classifiers yields the best performance. Also ‘majority voting’ was shown to perform a bit better than ‘average of probabilities’.

For the existing medical image edge detection algorithm image reconstruction accuracy is not high, the fitness of optimization coefficient is low, resulting in the detection results of low information recall, poor smoothness and low detection accuracy, we proposes an edge detection algorithm of cancer image based on deep learning. Firstly, the three-dimensional surface structure reconstruction model of cancer image was constructed. Secondly, the edge contour feature extraction method was used to extract the fine-grained features of cancer cells in the cancer image. Finally, the multi-dimensional pixel feature distributed recombination model of cancer image was constructed, and the fine-grained feature segmentation method was adopted to realize regional fusion and information recombination, and the ultra-fine particle feature was extracted. The adaptive optimization of edge detection was realized by combining with deep learning algorithm. The adaptive optimization in the process of edge detection was realized by combining with the deep learning algorithm. The experimental results show that the three-dimensional reconstruction accuracy of the proposed algorithm is about 95%, the fitness of the optimization coefficient is high, the algorithm has a strong edge information detection ability, and the output result smoothness and the accuracy of edge feature detection are high, which can effectively realize the detection of cancer image edge., GRAPHICAL ABSTRACT

Classification of human emotions based on electroencephalography (EEG) is a very popular topic nowadays in the provision of human health care and well-being. Fast and effective emotion recognition can play an important role in understanding a patient’s emotions and in monitoring stress levels in real-time. Due to the noisy and non-linear nature of the EEG signal, it is still difficult to understand emotions and can generate large feature vectors. In this article, we have proposed an efficient spatial feature extraction and feature selection method with a short processing time. The raw EEG signal is first divided into a smaller set of eigenmode functions called (IMF) using the empirical model-based decomposition proposed in our work, known as intensive multivariate empirical mode decomposition (iMEMD). The Spatio-temporal analysis is performed with Complex Continuous Wavelet Transform (CCWT) to collect all the information in the time and frequency domains. The multiple model extraction method uses three deep neural networks (DNNs) to extract features and dissect them together to have a combined feature vector. To overcome the computational curse, we propose a method of differential entropy and mutual information, which further reduces feature size by selecting high-quality features and pooling the k-means results to produce less dimensional qualitative feature vectors. The system seems complex, but once the network is trained with this model, real-time application testing and validation with good classification performance is fast. The proposed method for selecting attributes for benchmarking is validated with two publicly available data sets, SEED, and DEAP. This method is less expensive to calculate than more modern sentiment recognition methods, provides real-time sentiment analysis, and offers good classification accuracy.

Digital and paper based documents co-exist in our daily lives. Seamless integration of information from both sources is crucial for efficient knowledge management. This paper address the algorithm that can handle the detection of document so that it can be captured easily to convert it into a digital form for automatic integration of relevant information in electronic work flows. It uses the deep learning technique to provide a solution which is more generalized and flexible than other available solutions.

To effectively mine the contents embedded in web images, it is useful to classify the images into different types so that they can be fed to different procedures for detailed analysis. The authors herein propose a hierarchical algorithm for efficiently classifying web images into four classes. Their algorithm consists of two stages: the first stage extracts global features reflecting the distributions of color, edge and gradient, and uses a support vector machine (SVM) classifier for preliminary classification. Images assigned low confidence by the first stage classifier are processed by the second stage, which further extracts local texture features represented in the bag-of-words framework and uses another SVM classifier for final classification. In addition, they design two fusion strategies to train the second-stage classifier and generate the final prediction depending on the usage of local features in the second stage. To validate the effectiveness of proposed method, they built a database containing more than 55,000 images from various sources. On their test image set, they obtained an overall classification accuracy of 98.4% and the processing speed is over 27 fps on an Intel(R) Xeon(R) central processing unit (2.90 GHz).

Motivation Sputum in the trachea is hard to expectorate and detect directly for the patients who are unconscious, especially those in Intensive Care Unit. Medical staff should always check the condition of sputum in the trachea. This is time-consuming and the necessary skills are difficult to acquire. Currently, there are few automatic approaches to serve as alternatives to this manual approach. Results We develop an automatic approach to diagnose the condition of the sputum. Our approach utilizes a system involving a medical device and quantitative analytic methods. In this approach, the time-frequency distribution of respiratory sound signals, determined from the spectrum, is treated as an image. The sputum detection is performed by interpreting the patterns in the image through the procedure of preprocessing and feature extraction. In this study, 272 respiratory sound samples (145 sputum sound and 127 non-sputum sound samples) are collected from 12 patients. We apply the method of leave-one out cross-validation to the 12 patients to assess the performance of our approach. That is, out of the 12 patients, 11 are randomly selected and their sound samples are used to predict the sound samples in the remaining one patient. The results show that our automatic approach can classify the sputum condition at an accuracy rate of 83.5%. Availability and implementation The matlab codes and examples of datasets explored in this work are available at Bioinformatics online. Supplementary information Supplementary data are available at Bioinformatics online.

The difficulty of very high resolution image change detection lies in the low interclass separability between the changed class and the unchanged class. According to experiments, we found that this separability can be improved by mining the relationship contained in the training samples. Based on this observation, a supervised change detection approach is proposed in this paper based on relationship learning. The proposed approach begins with enriching the training samples based on their neighborhood relationship and label coherence; this relationship is then learned simultaneously with the classifier, and, finally, the latter classification performance benefits from the learned relationship. Experiments demonstrate the effectiveness of the proposed approach.

We address domain adaptation in the context of clustering where we are given a set of unlabeled data, coming from several domains, and the goal is to group data into different categories regardless of the domain they come from. This is a challenging problem since we do not have any supervision unlike most adaptation scenarios studied earlier, and is very relevant in practical industry applications where labeled data often comes at a premium especially while deploying services that do not have a comparable predecessor. Our philosophy in addressing this problem draws motivation from the concept of dirty paper coding, a communications technique where the signal being transmitted through a noisy channel is encoded with priors on the possible noise patterns to assist reliable decoding of the signal at the receiver. We focus on image applications in this paper, where we encode priors on possible image domain shift factors such as viewpoint, lighting, blur and appearance variations and utilize geometric adaptation mechanisms to perform clustering. We illustrate the utility of our approach on standard datasets involving objects and faces, by obtaining around 18% improvement on average over existing approaches.

Background Owing to the COVID-19 pandemic and the imminent collapse of health care systems following the exhaustion of financial, hospital, and medicinal resources, the World Health Organization changed the alert level of the COVID-19 pandemic from high to very high. Meanwhile, more cost-effective and precise COVID-19 detection methods are being preferred worldwide. Objective Machine vision–based COVID-19 detection methods, especially deep learning as a diagnostic method in the early stages of the pandemic, have been assigned great importance during the pandemic. This study aimed to design a highly efficient computer-aided detection (CAD) system for COVID-19 by using a neural search architecture network (NASNet)–based algorithm. Methods NASNet, a state-of-the-art pretrained convolutional neural network for image feature extraction, was adopted to identify patients with COVID-19 in their early stages of the disease. A local data set, comprising 10,153 computed tomography scans of 190 patients with and 59 without COVID-19 was used. Results After fitting on the training data set, hyperparameter tuning, and topological alterations of the classifier block, the proposed NASNet-based model was evaluated on the test data set and yielded remarkable results. The proposed model's performance achieved a detection sensitivity, specificity, and accuracy of 0.999, 0.986, and 0.996, respectively. Conclusions The proposed model achieved acceptable results in the categorization of 2 data classes. Therefore, a CAD system was designed on the basis of this model for COVID-19 detection using multiple lung computed tomography scans. The system differentiated all COVID-19 cases from non–COVID-19 ones without any error in the application phase. Overall, the proposed deep learning–based CAD system can greatly help radiologists detect COVID-19 in its early stages. During the COVID-19 pandemic, the use of a CAD system as a screening tool would accelerate disease detection and prevent the loss of health care resources.

Cardiotocography (CTG) is a standard tool for the assessment of fetal well-being during pregnancy and delivery. However, its interpretation is associated with high inter- and intra-observer variability. Since its introduction there have been numerous attempts to develop computerized systems assisting the evaluation of the CTG recording. Nevertheless these systems are still hardly used in a delivery ward. Two main approaches to computerized evaluation are encountered in the literature; the first one emulates existing guidelines, while the second one is more of a data-driven approach using signal processing and computational methods. The latter employs preprocessing, feature extraction/selection and a classifier that discriminates between two or more classes/conditions. These classes are often formed using the umbilical cord artery pH value measured after delivery. In this work an approach to Fetal Heart Rate (FHR) classification using pH is presented that could serve as a benchmark for reporting results on the unique open-access CTU-UHB CTG database, the largest and the only freely available database of this kind. The overall results using a very small number of features and a Least Squares Support Vector Machine (LS-SVM) classifier, are in accordance to the ones encountered in the literature and outperform the results of a baseline classification scheme proving the utility of using advanced data processing methods. Therefore the achieved results can be used as a benchmark for future research involving more informative features and/or better classification algorithms. Validerad;2017;Nivå 2;2017-12-05 (rokbeg)

Background: Instagram, with millions of posts per day, can be used to inform public health surveillance targets and policies. However, current research relying on image-based data often relies on hand coding of images which is time consuming and costly, ultimately limiting the scope of the study. Current best practices in automated image classification (e.g., support vector machine (SVM), Backpropagation (BP) neural network, and artificial neural network) are limited in their capacity to accurately distinguish between objects within images. Objective: This study demonstrates how convolutional neural network (CNN) can be used to extract unique features within an image and how SVM can then be used to classify the image. Methods: Images of waterpipes or hookah (an emerging tobacco product possessing similar harms to that of cigarettes) were collected from Instagram and used in analyses (n=840). CNN was used to extract unique features from images identified to contain waterpipes. A SVM classifier was built to distinguish between images with and without waterpipes. Methods for image classification were then compared to show how a CNN + SVM classifier could improve accuracy. Results: As the number of the validated training images increased, the total number of extracted features increased. Additionally, as the number of features learned by the SVM classifier increased, the average level of accuracy increased. Overall, 99.5% of the 420 images classified were correctly identified as either hookah or non-hookah images. This level of accuracy was an improvement over earlier methods that used SVM, CNN or Bag of Features (BOF) alone. Conclusions: CNN extracts more features of the images allowing a SVM classifier to be better informed, resulting in higher accuracy compared with methods that extract fewer features. Future research can use this method to grow the scope of image-based studies., Journal of Medical Internet Research

This paper presents a new contribution to the recommendation system following E-commerce reviews. Here, it implements a dictionary-based sentiment analysis based on four main steps (i) Pre-processing (ii) Semantic word extraction (iii) Feature Extraction (iv) Classification. Initially, the tweets or reviews extracted from the database are subjected to pre-processing, which involves three processes: stop word removal, stemming, and blank space removal. In the semantic word extraction process, the semantic words from the dictionary will be extracted by matching with the extracted keywords. The next process is the feature extraction, which extracts joint holoentropy and cross holoentropy of all keywords, and further proceeds with conditional holoentropy-based feature selection process. Finally, the selected features are subjected to a classification process using Deep Belief Network (DBN), which is a well-performing deep learning algorithm. Moreover, the main contribution of the paper relies on the improvement of DBN architecture, where the number of hidden neurons and activation function is optimally tuned by a hybrid optimization algorithm termed as Grey linked Chick Update-based Chicken Swarm Optimization (GCU-CSO). At last, the comparative analysis of proposed and conventional algorithms in terms of various performance metrics proves the efficacy of the proposed E-commerce recommendation system.

Emotion recognition based on electroencephalography (EEG) signal features is now one of the booming big data research areas. As the number of commercial EEG devices in the current market increases, there is a need to understand current trends and provide researchers and young practitioners with insights into future investigations of emotion recognition systems. This paper aims to evaluate popular consumer-grade EEG devices’ status and review relevant studies that examined the reliability of these low-cost devices for emotion recognition over the last five years. Additionally, a comparison with research-grade devices is conducted. This paper also highlights EEG-based emotion recognition research’s key areas, including different feature extraction capabilities, characteristics, and machine learning algorithms. Finally, the main challenges for building an EEG-based emotion recognition system, focusing on the data collection process with commercial EEG devices and machine learning algorithms’ performance, are presented.

The development of deep learning has rapidly updated image-based localization techniques. This paper presents a review and comparison of the current state-of-the-art methods for image-based localization using deep learning in the indoor environment. Traditional Global Structure from Motion (SfM) pipeline and learning-based pipeline from the recent techniques have been analyzed. Based on the pipeline, the methods are categorized into three groups: learned features and matching, learned relative pose estimation, and learned absolute pose estimation. Since multiple sensors are used in many applications, sensor fusion techniques including image information, have been briefly reviewed in this paper as well. Furthermore, the paper discusses challenges in these methods and concludes learned features and matching is the more competitive method for indoor localization.

Research work in the field of offline Arabic handwriting recognition has seen an exponential growth during past decades. Despite much work dedicated to the recognition of handwritten Arabic text, there remains a lot to achieve in this regard. Though there exists a lot of work that is confined to the Arabic text, this paper presents an approach towards classifying and recognizing text written in one of the famous scripts of Arabic language i.e. Kufic script. The approach based on character segmentation does not perform well in recognizing Kufic text due to various complexities. The proposed system is based on word segmentation and employs a Histogram of Oriented Gradient (HOG) and Local Binary Pattern (LBP) feature extraction techniques. Later in the paper a comparison is given between the numerical results of proposed technique with previous Arabic text recognition techniques to show the effectiveness of present work. This approach yields effective results with 97.05% accuracy in recognizing the Arabic text written in Kufic script using Polynomial kernel of SVM classifier. Experimental results show that the proposed system for recognition of Kufic script performs better than the previous recognition systems for Arabic text.

In this paper, a new approach to time series classification is proposed. It transforms the scalar time series into a two-dimensional space of amplitude (time series values) and a change of amplitude (increment). Subsequently, it uses this representation to plot the data. One figure is produced for each time series. In consequence, the time series classification problem is converted into the visual pattern recognition problem. This transformation allows applying a wide range of algorithms for standard pattern recognition – in this domain, there are more options to choose from than in the domain of time series classification. In this paper, we demonstrated the high effectiveness of the new method in a series of experiments on publicly available time series. We compare our results with several state-of-the-art approaches dedicated to time series classification. The new method is robust and stable. It works for time series of differing lengths and is easy to extend and alter. Even with a baseline variant presented in an empirical study in this paper, it achieves a satisfying classification accuracy. Furthermore, the proposed conversion of raw time series into images that are subjected to feature extraction opens the possibility to apply standard clustering algorithms.

This paper proposes a cluster-aware supervised learning (CluSL) framework, which integrates the clustering analysis with supervised learning. The objective of CluSL is to simultaneously find the best clusters of the data points and minimize the sum of loss functions within each cluster. This framework has many potential applications in healthcare, operations management, manufacturing, and so on. Because CluSL, in general, is nonconvex, we develop a regularized alternating minimization (RAM) algorithm to solve it, where at each iteration, we penalize the distance between the current clustering solution and the one from the previous iteration. By choosing a proper penalty function, we show that each iteration of the RAM algorithm can be computed efficiently. We further prove that the proposed RAM algorithm will always converge to a stationary point within a finite number of iterations. This is the first known convergence result in cluster-aware learning literature. Furthermore, we extend CluSL to the high-dimensional data sets, termed the F-CluSL framework. In F-CluSL, we cluster features and minimize loss function at the same time. Similarly, to solve F-CluSL, a variant of the RAM algorithm (i.e., F-RAM) is developed and proven to be convergent to an [Formula: see text]-stationary point. Our numerical studies demonstrate that the proposed CluSL and F-CluSL can outperform the existing ones such as random forests and support vector classification, both in the interpretability of learning results and in prediction accuracy. Summary of Contribution: Aligned with the mission and scope of the INFORMS Journal on Computing, this paper proposes a cluster-aware supervised learning (CluSL) framework, which integrates the clustering analysis with supervised learning. Because CluSL is, in general, nonconvex, a regularized alternating projection algorithm is developed to solve it and is proven to always find a stationary solution. We further generalize the framework to the high-dimensional data set, F-CluSL. Our numerical studies demonstrate that the proposed CluSL and F-CluSL can deliver more interpretable learning results and outperform the existing ones such as random forests and support vector classification in computational time and prediction accuracy.

Subtle facial expression recognition (SFER) aims to classify facial expressions with very low intensity into corresponding human emotions. Subtle facial expression can be regarded as a special kind of facial expression, whose facial muscle movements are more difficult to capture. In the last decade, various methods have been developed for common facial expression recognition (FER). However, most of them failed to automatically find the most discriminative parts of facial expression and the correlation of muscle movements when human makes facial expression, which makes them unsuitable for SFER. To better solve SFER problem, an attention mechanism based model focusing on salient local regions and their correlations is proposed in this paper. The proposed method: 1) utilizes multiple attention blocks to attend to distinct discriminative regions and extract corresponding local features automatically, 2) a correlation attention module is integrated in the model to extract global correlation feature over the salient regions, and finally 3) fuses the correlation feature and local features in an efficient way for the final facial expression classification. By this way, the useful but subtle local information can be utilized in more detail, and the correlation of different local regions is also extracted. Extensive experiment on the LSEMSW and CK+ datasets shows that the method proposed in this paper achieves superior results, which demonstrates its effectiveness.

With the development of the Internet, network security has aroused people’s attention. It can be said that a secure network environment is a basis for the rapid and sound development of the Internet. Phishing is an essential class of cybercriminals which is a malicious act of tricking users into clicking on phishing links, stealing user information, and ultimately using user data to fake logging in with related accounts to steal funds. Network security is an iterative issue of attack and defense. The methods of phishing and the technology of phishing detection are constantly being updated. Traditional methods for identifying phishing links rely on blacklists and whitelists, but this cannot identify new phishing links. Therefore, we need to solve how to predict whether a newly emerging link is a phishing website and improve the accuracy of the prediction. With the maturity of machine learning technology, prediction has become a vital ability. This paper offers a state-of-the-art survey on methods for phishing website detection. It starts with the life cycle of phishing, introduces common anti-phishing methods, mainly focuses on the method of identifying phishing links, and has an in-depth understanding of machine learning-based solutions, including data collection, feature extraction, modeling, and evaluation performance. This paper provides a detailed comparison of various solutions for phishing website detection.

Speaker recognition is a task of identifying persons from their voices. Recently, deep learning has dramatically revolutionized speaker recognition. However, there is lack of comprehensive reviews on the exciting progress. In this paper, we review several major subtasks of speaker recognition, including speaker verification, identification, diarization, and robust speaker recognition, with a focus on deep-learning-based methods. Because the major advantage of deep learning over conventional methods is its representation ability, which is able to produce highly abstract embedding features from utterances, we first pay close attention to deep-learning-based speaker feature extraction, including the inputs, network structures, temporal pooling strategies, and objective functions respectively, which are the fundamental components of many speaker recognition subtasks. Then, we make an overview of speaker diarization, with an emphasis of recent supervised, end-to-end, and online diarization. Finally, we survey robust speaker recognition from the perspectives of domain adaptation and speech enhancement, which are two major approaches of dealing with domain mismatch and noise problems. Popular and recently released corpora are listed at the end of the paper.

Human Face expression Recognition is one of the most powerful and challenging tasks in social communication. Generally, face expressions are natural and direct means for human beings to communicate their emotions and intentions. Face expressions are the key characteristics of non-verbal communication. This paper describes the survey of Face Expression Recognition (FER) techniques which include the three major stages such as preprocessing, feature extraction and classification. This survey explains the various types of FER techniques with its major contributions. The performance of various FER techniques is compared based on the number of expressions recognized and complexity of algorithms. Databases like JAFFE, CK, and some other variety of facial expression databases are discussed in this survey. The study on classifiers gather from recent papers reveals a more powerful and reliable understanding of the peculiar characteristics of classifiers for research fellows.

PurposePeriodic inspection and maintenance are essential for effective pavement preservation. Cracks not only affect the appearance of the road and reduce the levelness, but also shorten the life of road. However, traditional road crack detection methods based on manual investigations and image processing are costly, inefficiency and unreliable. The research aims to replace the traditional road crack detection method and further improve the detection effect.Design/methodology/approachIn this paper, a crack detection method based on matrix network fusing corner-based detection and segmentation network is proposed to effectively identify cracks. The method combines ResNet 152 with matrix network as the backbone network to achieve feature reuse of the crack. The crack region is identified by corners, and segmentation network is constructed to extract the crack. Finally, parameters such as the length and width of the cracks were calculated from the geometric characteristics of the cracks and the relative errors with the actual values were 4.23 and 6.98% respectively.FindingsTo improve the accuracy of crack detection, the model was optimized with the Adam algorithm and mixed with two publicly available datasets for model training and testing and compared with various methods. The results show that the detection performance of our method is better than many excellent algorithms, and the anti-interference ability is strong.Originality/valueThis paper proposed a new type of road crack detection method. The detection effect is better than a variety of detection algorithms and has strong anti-interference ability, which can completely replace traditional crack detection methods and meet engineering needs.

Surface defect detection of industrial products has become a promising area of research. Among the existing defect detection algorithms, most of the CNN-based methods can achieve the task of defect detection under ideal experimental conditions. However, the accuracy of defect detection is easily affected by the different lighting conditions of the environment and the inconsistency of the defect scale. Therefore, general deep learning methods have difficulties in solving the problem of defect detection in complex scenes. In this paper, a defect detection network based on low-light enhancement and multi-scale feature extraction (LE–MSFE–DDNet) is proposed. There are two blocks in the proposed network, including a low-light enhancement block and a SE-FP block. In the low-light enhancement block, the deep network is applied to enhance light adaptation of the deconstructed low-light feature map in our network. The influence of illumination inconsistency is weakened by the introduction of this block. In the SE-FP block, the dependencies between different channels are combined with the multi-scale feature extraction. The defects with different scales are accurately located through the combination of this block and our network. In addition, a Fine Cans Defect dataset based on the surface of fine cans is collected by this paper to verify the feasibility of the proposed network. The proposed model is compared with the state-of-the-art object detection network and the proposed method achieves 94.3% average accuracy on the Fine Cans Defect dataset. The experimental results show that the proposed method outperforms the state-of-the-art method for surface defect detection.

The application of deep learning techniques in hyperspectral satellite images (HSI) classification has led to a significant increase in accuracy compared to machine learning techniques. The research objective is now shifted toward gaining classification accuracy with less number of spectral bands and training samples. With this research objective, the recent development of the hybrid convolution neural network (CNN) has resulted in the improvement of the accuracy of land use land cover classification of hyperspectral satellite images (HSI) and also provided a new dimension to explore. This paper proposes a variant of the 3D-2D CNN Hybrid model to extract representational features from a different class of land use land cover using different receptive fields in a multi-stack arrangement. The proposed model focused on reducing the number of bands required for classification and training samples but on the other hand improving the diverse representational feature extraction. The paper also explores through experimentation the possible replacement of monotonic activation function ReLu using non-monotonic functions like Swish and Mish in 3D-2D CNN Hybrid models. The experiments are carried out using publically available HSI datasets with a special focus on small training datasets to establish the efficiency and robustness of the proposed model.

When the turning tool has worn and failed but the failure is not found, if it continues to be used for processing, it will break, and cause the workpiece to be scrapped, and even damage the machine tool. In order to avoid the loss caused by turning tool wear, the remaining useful life (RUL) prediction of turning tool wear has become a hot research topic in recent years. For RUL prediction in turning tools, the traditional machine is difficult to acquire sufficient degradation data and inconsistent data distribution among different turning tools in engineering, and they cannot provide better prediction accuracy to some extent. To solve the above problems, this paper proposes a multi-granularity feature extraction (MGFE) method based on the gray-level co-occurrence matrix (GLCM) and random forest (RF). Moreover, a health indicator (HI) of turning tools in the source domain was obtained. The common representative features in HI sequence of target domain was transferred to source domain and builds the condition monitoring and life prediction system of turning tools based on extreme learning machine and transfer learning. Finally, extreme vector machine (ELM) is used to construct the RUL prediction model. The research results show that the model constructed in this paper is effective in RUL prediction and can significantly improve the prediction accuracy of remaining useful life.

"Digital" agriculture is rapidly affecting the value of agricultural output. Robotic picking of the ripe agricultural product enables accurate and rapid picking, making agricultural harvesting intelligent. How to increase product output has also become a challenge for digital agriculture. During the cherry growth process, realizing the rapid and accurate detection of cherry fruits is the key to the development of cherry fruits in digital agriculture. Due to the inaccurate detection of cherry fruits, environmental problems such as shading have become the biggest challenge for cherry fruit detection. This paper proposes an improved YOLO-V4 deep learning algorithm to detect cherry fruits. This model is suitable for cherry fruits with a small volume. It is proposed to increase the network based on the YOLO-V4 backbone network CSPDarknet53 network, combined with DenseNet The density between layers, the a priori box in the YOLO-V4 model, is changed to a circular marker box that fits the shape of the cherry fruit. Based on the improved YOLO-V4 model, the feature extraction is enhanced, the network structure is deepened, and the detection speed is improved. To verify the effectiveness of this method, different deep learning algorithms of YOLO-V3, YOLO-V3-dense and YOLO-V4 are compared. The results show that the mAP (average accuracy) value obtained by using the improved YOLO-V4 model (YOLO-V4-dense) network in this paper is 0.15 higher than that of yolov4. In actual orchard applications, cherries with different ripeness of cherries in the same area can be detected, and the fruits with larger ripeness differences can be artificially intervened, and finally, the yield of cherry fruits can be increased.

Good prognostic health management (PHM) plays a crucial role in industrial production and other fields. The accurate prediction of remaining useful life (RUL) can ensure good working condition of machines, and the selection of health indicators (HI) is the key to prediction. Due to the length and noise of the data, the selection of features requires a lot of prior knowledge. Therefore, a novel convolution-based attention mechanism bidirectional long and short-term memory (CABLSTM) network is proposed to achieve the end-to-end lifetime prediction of rotating machinery in this paper. Unlike concatenating two networks, the model in this paper is used to convolute the cell states of Bi-LSTM. Firstly, the input signal is performed through CNN to obtain feature information. Secondly, the obtained features are fed into the Bi-LSTM network with attention mechanism for convolution operation to obtain time-frequency information to construct HI. Finally, the training data are normalized to predict RUL. Bi-LSTM can capture features in longer time-frequency information, and attention mechanism can give input influence weight, and highlight its effective characteristics to obtain better prediction accuracy. The complex process of feature extraction, HI construction, and RUL prediction is combined in one algorithm by deep learning. To verify the performance of the method in this paper, experiments were conducted on the bearing dataset of PRONOSTIA, and compared with other methods. The results showed that the method outperforms other methods due to its better accuracy and prediction precision.

Objective:Electroencephalography (EEG) has an influential role in neuroscience and commercial applications. Most of the tools available for EEG signal analysis use machine learning to extract the required information. So, the study of robust techniques for feature extraction and classification is an important thing to understand the practical use of EEG. The paper aims that if there is any special tool for a particular task. Which feature domain or classifier has a significant role in EEG signal analysis?Approach:It presents a detailed report of the current trend for bio-electrical signals classification focusing on various classifiers’ advantages and disadvantages. This study includes literature from 2000 to 2021 with a brief description of EEG signal origin and advancement in classification techniques.Results:Randomly used classifiers for EEG signal can be categorized into five classes, namely Linear Classifiers, Nearest Neighbor Classifiers, Nonlinear Bayesian Classifiers, Neural Networks, and Combinations of Classifiers. Approximately 40% of studies use Support Vector Machine, Nearest Neighbor, and their combination with others. For specific tasks, particular classifiers are recommended in the survey. Features can be defined into four categories, namely TDFs, FDFs, TFDFs, and statistical features, where 39% of studies used TFDFs. Multi-domains features are preferred when the required information cannot be obtained from one domain.Significance:The paper summarizes the recent approaches for feature extraction and classification of EEG signals. It describes the brain waves with their classification, related behavior, and task with the physiological correlation. The comparative analysis of different classifiers, toolbox, the channel used, accuracy, and the number of subjects from various studies can help the practitioners choose a suitable classifier. Furthermore, future directions can cope up with the relevant problems and can lead to accurate classification.

Accurate online classification of disturbance events in a transmission network is an important part of wide-area monitoring. Although many conventional machine learning techniques are very successful in classifying events, they rely on extracting information from PMU data at control centers and processing them through CPU/GPUs, which are highly inefficient in terms of energy consumption. To solve this challenge without compromising accuracy, this paper presents a novel methodology based on event-driven neuromorphic computing architecture for classification of power system disturbances. A Spiking Neural Network (SNN)-based computing framework is proposed, which exploits sparsity in disturbances and promotes local event driven operation for unsupervised learning and inference from incoming data. Spatio-temporal information of PMU signals is first extracted and encoded into spike trains and classification is achieved with SNN-based supervised and unsupervised learning framework. Moreover, a QR decomposition-based selection technique is proposed to identify signals participating in the low rank subspace of multiple disturbance events. Performance of the proposed method is validated on data collected from a 16-machine, 5-area New England-New York system., 11 pages, 8 figures. Paper has been accepted for publication in IEEE Transactions on Smart Grid on 26th of November, 2020. The link is as follows. https://ieeexplore.ieee.org/document/9290393/authors#authors

Facial Expression Recognition (FER) is an attractive and demanding problem in human computer interaction. Facial expressions are one of the most dominant and usual means for a human to communicate their feelings and intentions. This paper proposes the Enhanced Modified Decision Based Unsymmetric Trimmed Median Filter (EMDBUTMF) method which removes the noisy pixels from the face image. Also, this paper proposes the Local Directional Number (LDN) pattern, Dominant Gradient Local Ternary Pattern (DGLTP) descriptor for feature extraction and Support Vector Machine (SVM) classifier for classification. The histogram features are extracted from face images using the LDN and DGLTP descriptors. The experimental result shows that the proposed method gives an improved performance of 88% accuracy on two publicly available facial expression datasets JAFFE and CK.

Convolutional neural networks (CNN) have been widely used in various tasks, largely due to their ability to efficiently extract n-gram features for text analysis and document representation. In this paper, we intend to insight the CNN model regarding its capability on text analysis. Vanilla CNNs do have weaknesses when it comes to the representation and feature extraction. Duplicate filters are inevitable with vanilla CNNs, which reduces the discriminative power of the representations. In addition, the current pooling operations either limit the CNN to the local optimum (i.e., max pooling) or they do not consider the importance of all features (i.e., mean pooling). In this paper, we propose two modules for vanilla CNNs to overcome these shortcomings. The first equips the CNN with discriminative filters (distinct filters with maximised divergence) and the second provides the ability to comprehensively extract all salient features. Specifically, our model increases the discriminative power of the model by maximizing the distance between different filters, and a novel global pooling mechanism for feature extraction. Validation tests against state-of-the-art baselines on five benchmark classification datasets achieve the competitive performance of our proposed model. Furthermore, visualization on upgrade filters and pooling features verify our hypothesis that the proposed model can receive discriminative filters and salient features.

This paper uses an improved deep learning algorithm to judge the rationality of the design of landscape image feature recognition. The preprocessing of the image is proposed to enhance the data. The deficiencies in landscape feature extraction are further addressed based on the new model. Then, the two-stage training method of the model is used to solve the problems of long training time and convergence difficulties in deep learning. Innovative methods for zoning and segmentation training of landscape pattern features are proposed, which makes model training faster and generates more creative landscape patterns. Because of the impact of too many types of landscape elements in landscape images, traditional convolutional neural networks can no longer effectively solve this problem. On this basis, a fully convolutional neural network model is designed to perform semantic segmentation of landscape elements in landscape images. Through the method of deconvolution, the pixel-level semantic segmentation is realized. Compared with the 65% accuracy rate of the convolutional neural network, the fully convolutional neural network has an accuracy rate of 90.3% for the recognition of landscape elements. The method is effective, accurate, and intelligent for the classification of landscape element design, which better improves the accuracy of classification, greatly reduces the cost of landscape element design classification, and ensures that the technical method is feasible. This paper classifies landscape behavior based on this model for full convolutional neural network landscape images and demonstrates the effectiveness of using the model. In terms of landscape image processing, the image evaluation provides a certain basis.

In practical applications of the motor imagery-based brain–computer interface (BCI) system, the differences in electroencephalogram (EEG) signal manifestation and corresponding rhythm ranges in different individuals pose a significant challenge. The corresponding EEG features in different frequency bands differ; therefore, personalized screening must be conducted to obtain information that is conducive to the classification of EEG signals for different motor imageries. Also, in current BCI system, to obtain more information, multi-channel electrodes are often used to collect EEG signals, but also increasing the complexity of calculation. In this paper, a personalized feature extraction method based on filter bank and elastic net and a personalized channel selection based on Deep Belief Network to obtain a classification accuracy similar to or even higher than using all channels is proposed. Compared with the typically used feature extraction and classification algorithms, this method obtains higher calculation rates and recognition accuracy and provides a theoretical reference for the practical application of BCI systems. The shortcomings of the common spatial pattern (CSP) algorithm are addressed. The major contribution of this paper is the flexible screening of feature vectors and channels containing more classification information based on individual differences, thereby preventing the manual adjustment of specific frequency ranges in traditional CSP for performing feature extraction and avoiding inputting all channels. In the case study, the highest test accuracy reaches 86.94%.

In this paper, we review state-of-the-art approaches that apply signal processing (SP) and machine learning (ML) to automate the detection of Alzheimer’s disease (AD) and its prodromal stages. In the first part of the document, we describe the economic and social implications of the disease, traditional diagnosis techniques, and the fundaments of automated AD detection. Then, we present electroencephalography (EEG) as an appropriate alternative for the early detection of AD, owing to its reduced cost, portability, and non-invasiveness. We also describe the main time and frequency domain EEG features that are employed in AD detection. Subsequently, we examine some of the main studies of the last decade that aim to provide an automatic detection of AD and its previous stages by means of SP and ML. In these studies, brain data was acquired using multiple medical techniques such as magnetic resonance imaging, positron emission tomography, and EEG. The main aspects of each approach, namely feature extraction, classification model, validation approach, and performance metrics, are compiled and discussed. Lastly, a set of conclusions and recommendations for future research on AD automatic detection are drawn in the final section of the paper.

This paper proposes a new hybrid method for automatic detection and recognition of the presence/absence of a protective mask on human's face. It combines visual features extracted using Convolutional Neural Network (CNN) with image histograms that convey information about pixel intensity. Several pre-trained models for building feature extraction systems using a CNN and several types of image histograms are considered in this paper. We test our approach on the Medical Mask Dataset and perform cross-corpus analysis on two other databases named Masked Faces (MAFA) and Real-World Masked Face Dataset (RMFD). We demonstrate that the proposed hybrid method increases the Unweighted Average Recalls (UARs) of recognition of the presence/absence of a protective mask on human's face in comparison with traditional CNNs on the MAFA and RMFD databases by 0.96% and 1.32%, respectively. The proposed method can be generalized and used for other tasks of biometry, computer vision, machine learning and automatic face recognition.

To create a safe and secure cyberspace, many researchers have created and proposed Intrusion Detection Systems (IDSs). Different types of IDSs are either designed to address only a single kind of intrusion or multiple variants. In this paper, we proposed a new context-aware feature extraction method as a pre-processing step for Convolutional Neural Network (CNN)-based multiclass intrusion detection. Feature selection is also used to reduce the feature space and classification time. For attack-type classification, we used CNN which is generally well-known for its better classification ability on image recognition tasks. We have exploited this ability to develop an IDS that can identify different types of intrusions ranging from 4 to 12 types. In our study, we evaluated our proposed Context-aware Feature Extraction-based CNN IDS on different datasets with multiple classes where it was able to successfully improve classification accuracy compared to the models with no preprocessing steps and other existing methods. We carried out the performance evaluation and comparison using four different datasets, NSL-KDD, CICIDS2017, ADFA-LD, and ADFA-WD. These datasets have either host-based or network-based features, which is another area that we investigated in this paper.

This paper simply presents a state of the art survey among the machine learning based approaches for the Glioma classification. As Glioma classification is a very challenging task in the field of medical science and this task is well addressed and taken by the fraternity of machine learning experts, who are working day and night to devise automated approaches that can automate this whole process of Glioma accurate classification from the various medical imaging modalities like Magnetic resonance imaging (MRI), Computed tomography (CT) etc. Although present machine learning techniques offers an opportunity to come up with a highly accurate and automated Glioma classification approach, by performing fusion among the various medical imaging modalities as well as utilizing the various features derived from the multi-modality medical imaging data. This paper also proposed an efficient and accurate automated approach of Glioma classification for the comparison analysis. This proposed approach is based on the use of hybrid ensemble learning model and hybrid feature extraction method, which relies on the Discrete wavelet Decomposition (DWD), Central pixel Neighbourhood Binary pattern (CNBP) and GLRLM (Gray level run length Matrix) methods in order to classify the Glioma (type of mostly diagnosed brain tumors) into Low grade Glioma and High grade Glioma from the fused MRI sequences. Improved eXtreme Gradient Boosting classifier is the hybrid ensemble learning model, which is used in this paper for the first time along with the proposed hybrid texture feature extraction method. Further this proposed approach is compared with the already existing state of the art approaches, which are based on the various machine learning classifiers like Support vector machine (SVM), K-Nearest neighbor (KNN), Naive Bayes (NB) etc. and conventional feature extraction methods in order to present a comprehensive and practical comparison study. The proposed approach is evaluated on the balanced large size local dataset consisting of MRI images of low and high grade Glioma collected from the various MRI centers located in Madhya Pradesh, India as well as on the popular global datasets like, BRATS 2013 and BRATS 2015 with various MRI fusion combinations (T1 + T1C + T2 + Flair, T1 + T1C + T2, T1 + T1C + Flair, T1C + T2 + Flair etc.). The proposed approach employing Improved eXtreme Gradient Boosting ensemble model offers highest accuracy of above 90% on the local dataset with the fusion of T1C + T2 + Flair MRI sequences.

Intelligent fault diagnosis of bearings is an essential issue in the field of health management and the prediction of rotating machinery systems. The traditional bearing intelligent diagnosis algorithms based on the combination of feature extraction and classification for signal processing require high expert experience, which are time-consuming and lack universality. Compared with traditional methods, the convolutional neural network(CNN) can extract features automatically from the original vibration time-domain signal without any preprocessing. The accuracy of intelligent fault diagnosis can be improved by utilizing the multi-layer nonlinear mapping capability of deep convolutional neural networks. In order to realize the intelligent diagnosis and improve the recognition rate, this paper adopts the strategy of widening convolution kernels to obtain a larger receptive field and proposes a network design process pattern based on this idea, in addition, obtains the convolutional neural network with wide convolution kernels (WKCNN) model through experiments. Based on the time-domain vibration signal, this paper generates more input data through expansion and adopts the wide kernels of the first two convolutional layers to quickly extract features to improve efficiency. The smaller convolution kernels are used for multi-layer nonlinear mapping to deepen the network and improve detection accuracy. The results show that WKCNN performs well in accuracy, anti-noise, and timeliness compared with other diagnostic methods.

Person reidentification (Re-ID) aims to match observations of individuals across multiple nonoverlapping camera views. Recently, metric learning-based methods have played important roles in addressing this task. However, metrics are mostly learned in supervised manners, of which the performance relies heavily on the quantity and quality of manual annotations. Meanwhile, metric learning-based algorithms generally project person features into a common subspace, in which the extracted features are shared by all views. However, it may result in information loss since these algorithms neglect the view-specific features. Besides, they assume person samples of different views are taken from the same distribution. Conversely, these samples are more likely to obey different distributions due to view condition changes. To this end, this paper proposes an unsupervised cross-view metric learning method based on the properties of data distributions. Specifically, person samples in each view are taken from a mixture of two distributions: one models common prosperities among camera views and the other focuses on view-specific properties. Based on this, we introduce a shared mapping to explore the shared features. Meanwhile, we construct view-specific mappings to extract and project view-related features into a common subspace. As a result, samples in the transformed subspace follow the same distribution and are equipped with comprehensive representations. In this paper, these mappings are learned in an unsupervised manner by clustering samples in the projected space. Experimental results on five cross-view datasets validate the effectiveness of the proposed method.