Sparse Representation Classification Research Articles

A robust sparse representation algorithm based on adaptive joint dictionary

AbstractSparse representation based on dictionary construction and learning methods have aroused interests in the field of face recognition. Aiming at the shortcomings of face feature dictionary not ‘clean’ and noise interference dictionary not ‘representative’ in sparse representation classification model, a new method named as robust sparse representation is proposed based on adaptive joint dictionary (RSR‐AJD). First, a fast low‐rank subspace recovery algorithm based on LogDet function (Fast LRSR‐LogDet) is proposed for accurate low‐rank facial intrinsic dictionary representing the similar structure of human face and low computational complexity. Then, the Iteratively Reweighted Robust Principal Component Analysis (IRRPCA) algorithm is used to get a more precise occlusion dictionary for depicting the possible discontinuous interference information attached to human face such as glasses occlusion or scarf occlusion etc. Finally, the above Fast LRSR‐LogDet algorithm and IRRPCA algorithm are adopted to construct the adaptive joint dictionary, which includes the low‐rank facial intrinsic dictionary, the occlusion dictionary and the remaining intra‐class variant dictionary for robust sparse coding. Experiments conducted on four popular databases (AR, Extended Yale B, LFW, and Pubfig) verify the robustness and effectiveness of the authors’ method.

LASSO Homotopy-Based Sparse Representation Classification for fNIRS-BCI.

Brain-computer interface (BCI) systems based on functional near-infrared spectroscopy (fNIRS) have been used as a way of facilitating communication between the brain and peripheral devices. The BCI provides an option to improve the walking pattern of people with poor walking dysfunction, by applying a rehabilitation process. A state-of-the-art step-wise BCI system includes data acquisition, pre-processing, channel selection, feature extraction, and classification. In fNIRS-based BCI (fNIRS-BCI), channel selection plays a vital role in enhancing the classification accuracy of the BCI problem. In this study, the concentration of blood oxygenation (HbO) in a resting state and in a walking state was used to decode the walking activity and the resting state of the subject, using channel selection by Least Absolute Shrinkage and Selection Operator (LASSO) homotopy-based sparse representation classification. The fNIRS signals of nine subjects were collected from the left hemisphere of the primary motor cortex. The subjects performed the task of walking on a treadmill for 10 s, followed by a 20 s rest. Appropriate filters were applied to the collected signals to remove motion artifacts and physiological noises. LASSO homotopy-based sparse representation was used to select the most significant channels, and then classification was performed to identify walking and resting states. For comparison, the statistical spatial features of mean, peak, variance, and skewness, and their combination, were used for classification. The classification results after channel selection were then compared with the classification based on the extracted features. The classifiers used for both methods were linear discrimination analysis (LDA), support vector machine (SVM), and logistic regression (LR). The study found that LASSO homotopy-based sparse representation classification successfully discriminated between the walking and resting states, with a better average classification accuracy (p < 0.016) of 91.32%. This research provides a step forward in improving the classification accuracy of fNIRS-BCI systems. The proposed methodology may also be used for rehabilitation purposes, such as controlling wheelchairs and prostheses, as well as an active rehabilitation training technique for patients with motor dysfunction.

Improved Multiple Vector Representations of Images and Robust Dictionary Learning

Each sparse representation classifier has different classification accuracy for different samples. It is difficult to achieve good performance with a single feature classification model. In order to balance the large-scale information and global features of images, a robust dictionary learning method based on image multi-vector representation is proposed in this paper. First, this proposed method generates a reasonable virtual image for the original image and obtains the multi-vector representation of all images. Second, the same dictionary learning algorithm is used for each vector representation to obtain multiple sets of image features. The proposed multi-vector representation can provide a good global understanding of the whole image contour and increase the content of dictionary learning. Last, the weighted fusion algorithm is used to classify the test samples. The introduction of influencing factors and the automatic adjustment of the weights of each classifier in the final decision results have a significant indigenous effect on better extracting image features. The study conducted experiments on the proposed algorithm on a number of widely used image databases. A large number of experimental results show that it effectively improves the accuracy of image classification. At the same time, to fully dig and exploit possible representation diversity might be a better way to lead to potential various appearances and high classification accuracy concerning the image.

Corresponding keypoint constrained sparse representation three‐dimensional ear recognition via one sample per person

When only one sample per person (OSPP) is registered in the gallery, it is difficult for ear recognition methods to sufficiently and effectively reduce the search range of the matching features, thus resulting in low computational efficiency and mismatch problems. A 3D ear biometric system using OSPP is proposed to solve this problem. By categorising ear images by shape and establishing the corresponding relationship between keypoints from ear images and regions (regional cluster) on the directional proposals that can be arranged to roughly face the ear image, the corresponding keypoints are obtained. Then, ear recognition is performed by combining corresponding keypoints and a multi-keypoint descriptor sparse representation classification method. The experimental results conducted on the University of Notre Dame Collection J2 dataset yielded a rank-1 recognition rate of 98.84%; furthermore, the time for one identification operation shared by each gallery subject was 0.047 ms.

Machine learning-based automatic detection of novel coronavirus (COVID-19) disease.

The pandemic was announced by the world health organization coronavirus (COVID-19) universal health dilemma. Any scientific appliance which contributes expeditious detection of coronavirus with a huge recognition rate may be excessively fruitful to doctors. In this environment, innovative automation like deep learning, machine learning, image processing and medical image like chest radiography (CXR), computed tomography (CT) has been refined promising solution contrary to COVID-19. Currently, a reverse transcription-polymerase chain reaction (RT-PCR) test has been used to detect the coronavirus. Due to the moratorium period is high on results tested and huge false negative estimates, substitute solutions are desired. Thus, an automated machine learning-based algorithm is proposed for the detection of COVID-19 and the grading of nine different datasets. This research impacts the grant of image processing and machine learning to expeditious and definite coronavirus detection using CXR and CT medical imaging. This results in early detection, diagnosis, and cure for the accomplishment of COVID-19 as early as possible. Firstly, images are preprocessed by normalization to enhance the quality of the image and removing of noise. Secondly, segmentation of images is done by fuzzy c-means clustering. Then various features namely, statistical, textural, histogram of gradients, and discrete wavelet transform are extracted (92) and selected from the feature vector by principle component analysis. Lastly, k-NN, SRC, ANN, and SVM are used to make decisions for normal, pneumonia, COVID-19 positive patients. The performance of the system has been validated by the k (5) fold cross-validation technique. The proposed algorithm achieves 91.70% (k-Nearest Neighbor), 94.40% (Sparse Representation Classifier), 96.16% (Artificial Neural Network), and 99.14% (Support Vector Machine) for COVID detection. The proposed results show feature combination and selection improves the performance in 14.34 s with machine learning and image processing techniques. Among k-NN, SRC, ANN, and SVM classifiers, SVM shows more efficient results that are promising and comparable with the literature. The proposed approach results in an improved recognition rate as compared to the literature review. Therefore, the algorithm proposed shows immense potential to benefit the radiologist for their findings. Also, fruitful in prior virus diagnosis and discriminate pneumonia between COVID-19 and other pandemics.

Automated Recognition of Imagined Commands From EEG Signals Using Multivariate Fast and Adaptive Empirical Mode Decomposition Based Method

In this letter, a novel automated approach for recognizing imagined commands using multichannel electroencephalogram (MEEG) signals is presented. The multivariate fast and adaptive empirical mode decomposition method decomposes the MEEG signals into various modes. The slope domain entropy and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$L_1$</tex-math></inline-formula> -norm features are obtained from the modes of MEEG signals. The machine learning models such as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">k</i> -nearest neighbor, sparse representation classifier, and dictionary learning (DL) techniques are used for the imagined command classification tasks. The efficacy of the proposed approach is evaluated using MEEG from a public database as input signals. The proposed approach has achieved average accuracy values of 60.72, 59.73, and 58.78% using a DL model and selected features for left versus right, up versus down, forward versus backward based imagined command categorization tasks.

Superpixel-Based Brownian Descriptor for Hyperspectral Image Classification

Exploring effective spectral&#x2013;spatial feature extraction methods is one of the most focused problems in current hyperspectral image (HSI) classification research. However, complex spectral&#x2013;spatial structure characteristics in HSIs, e.g., shape-variable spatial structure and nonlinear spectral structure, are difficult to be effectively extracted and jointly represented. To overcome this issue, a novel superpixel-based Brownian descriptor (SBD) method for HSI classification is proposed in this article. In specific, superpixel segmentation is first used to extract shape-adaptive spatial structure information from dimension-reduced HSI, leading to generate nonoverlapping homogeneous 3-D image blocks. Then, similar pixels within the 3-D image block are jointly represented by a new local spectral&#x2013;spatial feature based on the Brownian descriptor (BD). This is the first time that the BD is introduced to measure both linear and nonlinear correlations among different spectral bands in HSI. On one hand, the integration of superpixel and BD helps to provide much richer and more valuable information for better discrimination between different categories. On the other hand, the SBD can effectively represent the internal structure characters within each 3-D image block of different spatial shapes by a symmetric positive definite matrix of a united form. Finally, considering that the SBD lies on the Riemannian manifold space, a log-Euclidean kernel sparse representation (LKSR) classifier is introduced to obtain the classification results. Experimental results on three widely used real hyperspectral datasets indicate the performance superiority of the proposed SBD method over several state-of-the-art techniques.

In-Bed Posture Classification Based on Sparse Representation in Redundant Dictionaries

Non-orthogonal signal representation using redundant dictionaries gradually gained popularity over the last decades. Sparse methods find major application in signal denoising, audio declipping, time-frequency analysis, and classification, to name a few. This paper is inspired by the exceptional results of sparse representation classification originally suggested for face recognition. We compare the method to other common classifiers using simulated as well as real datasets. In the latter the proposed method is tested with real pressure data from a bed equipped with a matrix of 30 × 11 pressure sensors. Here the method outperforms standard classification methods (surpassing 91 % accuracy) without need of parameter selection or special user's skills. Furthermore it offers a means of dealing with occlusions, whose results are presented as well.

Phonetic Segmentation using a Wavelet-based Speech Cepstral Features and Sparse Representation Classifier

Speech segmentation is the process of dividing speech signal into distinct acoustic blocks that could be words, syllables or phonemes. Phonetic segmentation is about finding the exact boundaries for the different phonemes that composes a specific speech signal. This problem is crucial for many applications, i.e. automatic speech recognition (ASR). In this paper we propose a new model-based text independent phonetic segmentation method based on wavelet packet speech parametrization features and using the sparse representation classifier (SRC). Experiments were performed on two datasets, the first is an English one derived from TIMIT corpus, while the second is an Arabic one derived from the Arabic speech corpus. Results showed that the proposed wavelet packet de composition features outperform the MFCC features in speech segmentation task, in terms of both F1-score and R-measure on both datasets. Results also indicate that the SRC gives higher hit rate than the famous k-Nearest Neighbors (k-NN) classifier on TIMIT dataset.

Ultrasound Image Texture Feature Learning-Based Breast Cancer Benign and Malignant Classification.

The use of ultrasound images to acquire breast cancer diagnosis information without invasion can reduce the physical and psychological pain of breast cancer patients and is of great significance for the diagnosis and treatment of breast cancer. There are some differences in the texture of breast cancer between benign and malignant cases. Therefore, this paper proposes an adaptive learning method based on ultrasonic image texture features to identify breast cancer. Specifically, firstly, we used dictionary learning and sparse representation to learn the ultrasonic image texture dictionary of benign and malignant cases, respectively, and then used the combination of the two dictionaries to represent the test image to obtain the texture distribution characteristics of the test image under the two dictionary representations, which called the sparse representation coefficient. Finally, these above features were filtered by sparse representation and sent to sparse representation classifier to establish benign and malignant classification model. 128 cases were randomly divided into training and testing sets according to 2: 1 for training and testing. The proposed method has achieved state-of-the-art results, with an accuracy of 0.9070 and the area under the receiver operating characteristic curve of 0.9459. The results demonstrate that the proposed method has the potential to be used in the clinical diagnosis of benign and malignant breast cancer.

Classification method of marine target motion pattern based on spatial-temporal trajectories

Classifying the motion pattern of marine targets is of important significance to promote target surveillance and management efficiency of marine area and to guarantee sea route safety. This paper proposes a moving target classification algorithm model based on channel extraction-segmentation-LCSCA-lp norm minimization. The algorithm firstly analyzes the entire distribution of channels in specific region, and defines the categories of potential ship motion patterns; on this basis, through secondary segmentation processing method, it obtains several line segment trajectories as training sample sets, to improve the accuracy of classification algorithm; then, it further uses the Leastsquares Cubic Spline Curves Approximation (LCSCA) technology to represent the training sample sets, and builds a motion pattern classification sample dictionary; finally, it uses lp norm minimized sparse representation classification model to realize the classification of motion patterns. The verification experiment based on real spatial-temporal trajectory dataset indicates that, this method can effectively realize the motion pattern classification of marine targets, and shows better time performance and classification accuracy than other representative classification methods.

Semantic understanding based on multi-feature kernel sparse representation and decision rules for mangrove growth

With the rapid development of remote sensing technology, using remote sensing technology is an important means to monitor the dynamic change of land cover and ecology. In view of the complexity of mangrove ecological monitoring in Dongzhaigang, Hainan Province of China, we propose a semantic understanding method of mangrove remote sensing image by combining a multi-feature kernel sparse classifier with a decision rule model in this paper. First, on the basis of multi-feature extraction, we take into account the spatial context relations of the samples and introduce the kernel function into the sparse representation classifier, a multi-feature kernel sparse representation classifier can be constructed to classify cover types of mangroves and their surrounding objects. Second, in view of growth conditions of mangrove area, we put forward a semantic understanding method of mangrove remote sensing image based on decision rules and divide mangrove and non-mangrove areas by combining classification results of the multi-feature kernel sparse representation classifier. We make a divisibility analysis based on the extracted features of spatial and spectral domains. Then select the best split attribute based on the maximum information gain criterion, to generate a semantic tree and extract semantic rules. Finally, we work on the semantic understanding of mangrove areas in line with decision rules and further divide mangrove areas into two categories: excellent growth and poor growth. Experimental results show that the proposed method can effectively identify mangrove areas and make decisions on mangrove growth.

Active Sonar Target Classification Method Based on Fisher’s Dictionary Learning

The marine environment is complex and changeable, and the interference of noise and reverberation seriously affects the classification performance of active sonar equipment. In particular, when the targets to be measured have similar characteristics, underwater classification becomes more complex. Therefore, a strong, recognizable algorithm needs to be developed that can handle similar feature targets in a reverberation environment. This paper combines Fisher’s discriminant criterion and a dictionary-learning-based sparse representation classification algorithm, and proposes an active sonar target classification method based on Fisher discriminant dictionary learning (FDDL). Based on the learning dictionaries, the proposed method introduces the Fisher restriction criterion to limit the sparse coefficients, thereby obtaining a more discriminating dictionary; finally, it distinguishes the category according to the reconstruction errors of the reconstructed signal and the signal to be measured. The classification performance is compared with the existing methods, such as SVM (Support Vector Machine), SRC (Sparse Representation Based Classification), D-KSVD (Discriminative K-Singular Value Decomposition), and LC-KSVD (label-consistent K-SVD), and the experimental results show that FDDL has a better classification performance than the existing classification methods.

A differential privacy-based classification system for edge computing in IoT

The blooming Internet of Things (IoTs) devices have brought many new types of sensing applications and methods to the traditional cloud-enabled IoTs framework. Hence, the new network framework becomes bidirectional gradually, in which IoTs devices can also perform moderate computation tasks instead of being solely used as data harvesters. This new coming framework known as edge computing significantly improves the traditional cloud-enabled network latency and dependency by shifting part of computation back to “local”. However, new security risks emerge when the edge computing shifts data and models back to the IoTs devices.Acies, a differential privacy based privacy-preserving classification system for edge computing is proposed to secure the classification models offloaded to edge devices. Acies supports popular classifiers such as Nearest Neighborhood, Support Vector Machine and Sparse Representation Classifier with a variety of feature selection methods. According to our evaluation on different datasets, classification models with Acies can be private and maintain high utility. Acies achieves reliable privacy protection under reconstruction attacks with minimal impact on classification accuracy (2% ∽ 5%) only. Acies outperforms the naive input dataset perturbation methods by up to 30% higher classification accuracy when the privacy requirements of the applications is high (the privacy budget ε is less than 2).

Review on engine vibration fault analysis based on data mining

Through equipment monitoring, the uptimes of machines are enhanced in the industrial applications. The unpredicted failures risks are minimized by the proper equipment monitoring. The machine vibrations are increased caused by the failure modes. The vibration data requires effective analysis by the accurate assessment of the machine equipment. For fault feature selection and detection of faults in rotating equipment, the empirical knowledge is required. Low efficiency of the methods and motor speed control are the main drawbacks of the existing techniques. So the basic aim of this paper is the detection of rotating equipment faults by utilizing the vibration analysis. The motor vibration is analyzed and monitored using spectrum analysis. The spectral content are extracted and fed into the classifier like k-Nearest neighbors (KNN), back-propagation neural network BPNN, Sparse Representation Classifier (SRC), Support vector machine (SVM) and Random Forest (RF) for the type of failure prediction and analyze the unbalance condition (UNB), bearing faults (BDF), and broken rotor bars (BRB) faults. The RF classifier is better as compared to other classifiers in terms of accuracy, precision and recalls values by approximately 10.92 %, 11.03 % and 20.13 % respectively.

Method of Metal Button Defect Detection Based on Extreme Learning Machine and Sparse Representation

We studied the algorithms of button defect detection, while metal button defect detection is seldom explored in the literature. In this paper, we propose to effectively improve the detecting speed, accuracy, and robustness of the model. We propose a new method of cascading Extreme Learning Machine (ELM) and Sparse Representation Classification (SRC). This method transforms the defect detection problem into a pattern classification problem. First, we preprocess the input button images via eliminating reflection, edges extraction, and dimensionality reduction. ELM has a faster learning speed and better detecting accuracy than the single layer perceptron and support layer machine. ELM was utilized to estimate the probability of the defective buttons, with parameters obtained from the five-fold cross-validation method. SRC was used to reclassify those button images with high noise. Our method improved the detection accuracy, as well as guaranteed the detecting speed. We showed state-of-the-art performance in comparison to other approaches.

Aircraft Engine Gas-Path Monitoring and Diagnostics Framework Based on a Hybrid Fault Recognition Approach

Considering the importance of continually improving the algorithms in aircraft engine diagnostic systems, the present paper proposes and benchmarks a gas-path monitoring and diagnostics framework through the Propulsion Diagnostic Methodology Evaluation Strategy (ProDiMES) software developed by NASA. The algorithm uses fleet-average and individual engine baseline models to compute feature vectors that form a fault classification with healthy and faulty engine classes. Using this classification, a hybrid fault-recognition technique based on regularized extreme learning machines and sparse representation classification was trained and validated to perform both fault detection and fault identification as a common process. The performance of the system was analyzed along with the results of other diagnostic frameworks through four stages of comparison based on different conditions, such as operating regimes, testing data, and metrics (detection, classification, and detection latency). The first three stages were devoted to the independent algorithm development and self-evaluation, while the final stage was related to a blind test case evaluated by NASA. The comparative analysis at all stages shows that the proposed algorithm outperforms all other diagnostic solutions published so far. Considering the advantages and the results obtained, the framework is a promising tool for aircraft engine monitoring and diagnostic systems.

Cooperative classification method for multi-fault diagnosis of machinery based on parameterized wavelet kernel extreme learning and sparse representation

Owing to complicated manual design and feature extraction processes of mechanical multi-fault classification and recognition methods, single-class classifiers do not adapt well to the high dimensionality and high self-similarity among mechanical vibration signals with nonlinear characteristics, resulting in insufficient classification accuracy. Combining the adaptive learning and feature extraction capabilities of the wavelet kernel-based extreme learning machine (WKELM) with the sparse representation classifier (SRC), this paper proposes a hybrid artificial bee colony (ABC)-optimized Morlet wavelet kernel-based extreme learning machine (ABC-MWKELM) with SRC. The two-level classifier is capable of cooperatively classifying multiple mechanical faults. First, a Morlet wavelet kernel-based extreme learning machine (MWKELM) model that matches the kernel function construction condition is presented. An ABC optimization algorithm is adopted to optimize critical parameters of the MWKELM model, thereby improving the accuracy and generalization ability of the model. In addition, the ABC algorithm is used to optimize the MWKELM model as a pre-classifier for classifying test samples. According to the posterior probability of each training sample category output, the threshold decision method selects candidate categories with a high confidence coefficient. When the preset meets the threshold, the classification result of the ABC-optimized MWKELM is directly adopted. Otherwise, the candidate training samples are used to construct a dictionary and the SRC is used to classify the test samples further. The experimental results demonstrate that compared with the ELM, WKELM, (ELM-SRC), and radial basis function (RBF)-KELM and SRC (RBFKELM-SRC), the method presented in this paper has significant advantages in terms of accuracy of mechanical multi-fault classification. The validity and applicability of this method in mechanical multi-fault classification were verified.

A Domain Adaptation Sparse Representation Classifier for Cross-Domain Electroencephalogram-Based Emotion Classification.

The brain-computer interface (BCI) interprets the physiological information of the human brain in the process of consciousness activity. It builds a direct information transmission channel between the brain and the outside world. As the most common non-invasive BCI modality, electroencephalogram (EEG) plays an important role in the emotion recognition of BCI; however, due to the individual variability and non-stationary of EEG signals, the construction of EEG-based emotion classifiers for different subjects, different sessions, and different devices is an important research direction. Domain adaptation utilizes data or knowledge from more than one domain and focuses on transferring knowledge from the source domain (SD) to the target domain (TD), in which the EEG data may be collected from different subjects, sessions, or devices. In this study, a new domain adaptation sparse representation classifier (DASRC) is proposed to address the cross-domain EEG-based emotion classification. To reduce the differences in domain distribution, the local information preserved criterion is exploited to project the samples from SD and TD into a shared subspace. A common domain-invariant dictionary is learned in the projection subspace so that an inherent connection can be built between SD and TD. In addition, both principal component analysis (PCA) and Fisher criteria are exploited to promote the recognition ability of the learned dictionary. Besides, an optimization method is proposed to alternatively update the subspace and dictionary learning. The comparison of CSFDDL shows the feasibility and competitive performance for cross-subject and cross-dataset EEG-based emotion classification problems.

Atom specific multiple kernel dictionary based Sparse Representation Classifier for medium scale image classification

Kernel based Sparse Representation Classifier (KSRC) can classify images with acceptable performance. In addition, Multiple Kernel Learning based SRC (MKL-SRC) computes the weighted sum of multiple kernels in order to construct a unified kernel while the weight of each kernel is calculated as a fixed value in the training phase. In this paper, an MKL-SRC with non-fixed kernel weights for dictionary atoms is proposed. Kernel weights are embedded as new variables to the main KSRC goal function and the resulted optimization problem is solved to find the sparse coefficients and kernel weights simultaneously. As a result, an atom specific multiple kernel dictionary is computed in the training phase which is used by SRC to classify test images. Also, it is proved that the resulting optimization problem is convex and is solvable via common algorithms. The experimental results demonstrate the effectiveness of the proposed approach.