Extreme Learning Machine Classifier Research Articles

Facial Expression Recognition of Instructor Using Deep Features and Extreme Learning Machine.

Classroom communication involves teacher's behavior and student's responses. Extensive research has been done on the analysis of student's facial expressions, but the impact of instructor's facial expressions is yet an unexplored area of research. Facial expression recognition has the potential to predict the impact of teacher's emotions in a classroom environment. Intelligent assessment of instructor behavior during lecture delivery not only might improve the learning environment but also could save time and resources utilized in manual assessment strategies. To address the issue of manual assessment, we propose an instructor's facial expression recognition approach within a classroom using a feedforward learning model. First, the face is detected from the acquired lecture videos and key frames are selected, discarding all the redundant frames for effective high-level feature extraction. Then, deep features are extracted using multiple convolution neural networks along with parameter tuning which are then fed to a classifier. For fast learning and good generalization of the algorithm, a regularized extreme learning machine (RELM) classifier is employed which classifies five different expressions of the instructor within the classroom. Experiments are conducted on a newly created instructor's facial expression dataset in classroom environments plus three benchmark facial datasets, i.e., Cohn–Kanade, the Japanese Female Facial Expression (JAFFE) dataset, and the Facial Expression Recognition 2013 (FER2013) dataset. Furthermore, the proposed method is compared with state-of-the-art techniques, traditional classifiers, and convolutional neural models. Experimentation results indicate significant performance gain on parameters such as accuracy, F1-score, and recall.

Detection and recognition of text traffic signs above the road

Based on the similarity between traffic sign images in the source and target domains, the parameters migrated from the source domain are utilised as the initial parameters of the faster region convolutional neural network (Faster-R-CNN), which is trained for detecting text traffic signs, and then fine-tune the network parameters based on the samples in target domain for obtaining the final network parameters. Moreover, the converted traffic sign images from RGB to HSV space is also used as the training samples of the network, thereby overcoming the under-learning problem of model caused by less training samples. The traditional efficient and accurate scene text (EAST) detection network model is tailored and a new recognition model is proposed based on the extreme learning machine (ELM) classifier to identify and classify the detected text traffic signs above the road. Experimental results in the natural scene demonstrate the effectiveness of the proposed method.

Detection and recognition of text traffic signs above the road

Based on the similarity between traffic sign images in the source and target domains, the parameters migrated from the source domain are utilised as the initial parameters of the faster region convolutional neural network (Faster-R-CNN), which is trained for detecting text traffic signs, and then fine-tune the network parameters based on the samples in target domain for obtaining the final network parameters. Moreover, the converted traffic sign images from RGB to HSV space is also used as the training samples of the network, thereby overcoming the under-learning problem of model caused by less training samples. The traditional efficient and accurate scene text (EAST) detection network model is tailored and a new recognition model is proposed based on the extreme learning machine (ELM) classifier to identify and classify the detected text traffic signs above the road. Experimental results in the natural scene demonstrate the effectiveness of the proposed method.

Facial Expression Recognition Using Dynamic Local Ternary Patterns With Kernel Extreme Learning Machine Classifier

Rapid growth in advanced human-computer interaction (HCI) based applications has led to the immense popularity of facial expression recognition (FER) research among computer vision and pattern recognition researchers. Lately, a robust texture descriptor named Dynamic Local Ternary Pattern (DLTP) developed for face liveness detection has proved to be very useful in preserving facial texture information. The findings motivated us to investigate DLTP in more detail and examine its usefulness in the FER task. To this end, a FER pipeline is developed, which uses a sequence of steps to detect possible facial expressions in a given input image. Given an input image, the pipeline first locates and registers faces in it. In the next step, using an image enhancement operator, the FER pipeline enhances the facial images. Afterward, from the enhanced images, facial features are extracted using the DLTP descriptor. Subsequently, the pipeline reduces dimensions of the high-dimensional DLTP features via Principal Component Analysis (PCA). Finally, using the multi-class Kernel Extreme Learning Machine (K-ELM) classifier, the proposed FER scheme classifies the features into facial expressions. Extensive experiments performed on four in-the-lab and one in-the-wild FER datasets confirmed the superiority of the method. Besides, the cross-dataset experiments performed on different combinations of the FER datasets revealed its robustness. Comparison results with several state-of-the-art FER methods demonstrate the usefulness of the proposed FER scheme. The pipeline with a recognition accuracy of 99.76%, 99.72%, 93.98%, 96.71%, and 78.75%, respectively, on the CK+, RaF, KDEF, JAFFE, and RAF-DB datasets, outperformed the previous state-of-the-art.

Data and Knowledge Twin Driven Integration for Large-Scale Device-Free Localization

Device-free localization (DFL) is becoming one of the attractive techniques in wireless sensing field, due to its advantage that the target does not need to be attached to any electronic device. However, most of the existed approaches for DFL can only obtain satisfactory localization performance in specific small area, they cannot function well when implemented to complex and large area. In order to tackle this issue, in this article, a hierarchical framework is developed for large-scale DFL based on data and knowledge twin driven integration, which consists of two phases, including the offline training phase and the online localization phase. In the offline training phase, the complex and large monitoring area is first divided into some subdomains by the K-means clustering algorithm, and then training corresponding number of broad learning (BL)-based DFL models for each subdomain using the Fresnel phase difference as the fingerprints. Meanwhile, a class-specific cost regulation extreme learning machine (CCR-ELM) classifier is also trained for determining the attribution of the reference points, which can alleviate the impacts of imbalanced data distribution on classification results. In the online localization phase, the attributions of the testing points are first judged through the trained CCR-ELM classifier, after that, estimating the target's location in the corresponding subdomains using BL-based DFL models. The validity of the proposed hierarchical framework is evaluated both in small and larger areas, respectively.

Classification of COVID-19 CT Scans via Extreme Learning Machine

Here, we use multi-type feature fusion and selection to predict COVID-19 infections on chest computed tomography (CT) scans. The scheme operates in four steps. Initially, we prepared a database containing COVID-19 pneumonia and normal CT scans. These images were retrieved from the Radiopaedia COVID-19 website. The images were divided into training and test sets in a ratio of 70:30. Then, multiple features were extracted from the training data. We used canonical correlation analysis to fuse the features into single vectors; this enhanced the predictive capacity. We next implemented a genetic algorithm (GA) in which an Extreme Learning Machine (ELM) served to assess GA fitness. Based on the ELM losses, the most discriminatory features were selected and saved as an ELM Model. Test images were sent to the model, and the best-selected features compared to those of the trained model to allow final predictions. Validation employed the collected chest CT scans. The best predictive accuracy of the ELM classifier was 93.9%; the scheme was effective.

Improving the Performance of Machine Learning-Based Network Intrusion Detection Systems on the UNSW-NB15 Dataset.

Networks are exposed to an increasing number of cyberattacks due to their vulnerabilities. So, cybersecurity strives to make networks as safe as possible, by introducing defense systems to detect any suspicious activities. However, firewalls and classical intrusion detection systems (IDSs) suffer from continuous updating of their defined databases to detect threats. The new directions of the IDSs aim to leverage the machine learning models to design more robust systems with higher detection rates and lower false alarm rates. This research presents a novel network IDS, which plays an important role in network security and faces the current cyberattacks on networks using the UNSW-NB15 dataset benchmark. Our proposed system is a dynamically scalable multiclass machine learning-based network IDS. It consists of several stages based on supervised machine learning. It starts with the Synthetic Minority Oversampling Technique (SMOTE) method to solve the imbalanced classes problem in the dataset and then selects the important features for each class existing in the dataset by the Gini Impurity criterion using the Extremely Randomized Trees Classifier (Extra Trees Classifier). After that, a pretrained extreme learning machine (ELM) model is responsible for detecting the attacks separately, “One-Versus-All” as a binary classifier for each of them. Finally, the ELM classifier outputs become the inputs to a fully connected layer in order to learn from all their combinations, followed by a logistic regression layer to make soft decisions for all classes. Results show that our proposed system performs better than related works in terms of accuracy, false alarm rate, Receiver Operating Characteristic (ROC), and Precision-Recall Curves (PRCs).

Clothing Style Recognition Method Based on Digital Image Processing

The use of digital image processing technology to identify clothing styles from clothing images has broad application potential in clothing consumption analysis, auxiliary clothing design and identity recognition. In the current clothing style recognition field for feature extraction and classification of clothing contours, the main methods are extreme learning machine classification based on wavelet Fourier descriptor, and Euclidean distance classification based on fusion features. However, these methods also have some shortcomings. Currently, there is still no effective method to extract features and classify clothing contours. In order to solve this problem, this paper is based on the SVM classification method of Fourier contour description features. The contour curvature characteristic point in the figure can be described as the bending characteristic of the clothing contour curve, and its calculation is simple and intuitive, which improves the judgment effect of the similarity performance of the Hausdorff distance. This paper constructs a multi-resolution framework of the new image, uses edge detection algorithms to obtain the edges of each level of resolution image, and expands them into narrow edge bands, and combines the point distribution model with the edge narrow bands obtained by edge detection to perform multi-resolution image ASM search for. This paper proposes an image fusion algorithm based on algebraic multigrid and adaptive block. Because the coarse grid extracted by the algebraic multi-grid method can extract the detailed information of the image to a certain extent, the original image can be reconstructed from the grid data. Experimental research shows that from the overall recognition rate of various clothing styles, the Fourier descriptor has a high recognition rate for each clothing, and the stability of multiple experiments is also good.

PSO-ELM with Time-varying Inertia Weight for Classification of SMILES Codes

Classification of active compounds is necessary to trivialize their properties and activities and can be carried out by identifying functional groups and their structures based on the simplified molecular input line entry system (SMILES) code. This study classified active compounds based on the SMILES structure using a combination of the machine learning techniques, extreme learning machine (ELM), and particle swarm optimization (PSO). ELM facilitates learning very quickly and at with high predictive accuracy. The PSO covers the limitations of the ELM, i.e., randomly determining the weight and bias, trial, and error in determining the number of hidden neurons. PSO optimizes the weight, bias, and number of hidden neurons. This study’s results indicate that the classification of active compounds’ function based on the SMILES code can be done by the ELM and PSO-ELM with high accuracy. The PSO-ELM can improve the accuracy performance of the ELM classification by optimizing the weight, bias, and number of hidden neurons automatically. Moreover, the proposed PSO-ELM is better than the PSO-ELM for average accuracy, computation time, and standard deviation. The proposed algorithm has the highest average accuracy compared to the ELM by increasing the accuracy to 2.54%, 6.43%, and 3.85% for 2, 3, and 4 classes, respectively. A comparison with other machine algorithms shows that the proposed is superior.

Building change detection from multi-source remote sensing images based on multi-feature fusion and extreme learning machine

ABSTRACT In this paper, in order to improve the accuracy of multi-source remote sensing image building change detection, we propose a method based on multi-feature fusion (MFF) and extreme learning machine (ELM) training. Firstly, we fuse the spectral feature difference image (DI) and textural feature (grey level co-occurrence matrix) DI obtained by change vector analysis (CVA), morphological building index DI, and shape feature DI obtained by subtraction to construct the final DI. Secondly, the coarse change detection map obtained by selecting a threshold for the DI saliency map obtained by the use of the frequency-domain significance (FDS) method is pre-classified by the fuzzy c-means (FCM) clustering algorithm. Finally, the neighborhood features obtained from the original images and the feature images of the changed pixels (buildings) and the unchanged pixels in the coarse change map are extracted and used as reliable samples for the ELM training. By using the trained ELM classifier, undetermined pixels are further separated into changed and unchanged classes. Finally, we combine the ELM classification result and the preclassification result together to form the final building change map. Experiments on two real multi-source datasets show that the proposed method can result in a significant improvement in multi-source remote sensing image building change detection performance.

Prediction of COVID-19 - Pneumonia based on Selected Deep Features and One Class Kernel Extreme Learning Machine

In this work, we propose a deep learning framework for the classification of COVID-19 pneumonia infection from normal chest CT scans. In this regard, a 15-layered convolutional neural network architecture is developed which extracts deep features from the selected image samples – collected from the Radiopeadia. Deep features are collected from two different layers, global average pool and fully connected layers, which are later combined using the max-layer detail (MLD) approach. Subsequently, a Correntropy technique is embedded in the main design to select the most discriminant features from the pool of features. One-class kernel extreme learning machine classifier is utilized for the final classification to achieving an average accuracy of 95.1%, and the sensitivity, specificity & precision rate of 95.1%, 95%, & 94% respectively. To further verify our claims, detailed statistical analyses based on standard error mean (SEM) is also provided, which proves the effectiveness of our proposed prediction design.

Extreme Learning Machine Based on Maximum Weighted Mean Discrepancy for Unsupervised Domain Adaptation

Extreme Learning Machine (ELM) has shown fast learning speed and good generalization property in single-domain problems, such as classification and regression. However, the assumption that the training and testing data are subject to identical distribution often leads to significant performance degradation of ELM in cross-domain problems. To cope with unsupervised domain adaptation problems by ELM, we propose a novel method called Extreme Learning Machine based on Maximum Weighted Mean Discrepancy (ELM-MWMD) in this paper, which learns an adaptive ELM classifier with both labeled source data and unlabeled target data. Firstly, the cross-domain weight coefficients are specifically designed and assigned for each sample in source and target domains, fully considering the effects of individual information. Then the source classifier is adapted to the target domain by minimizing the distribution discrepancy between the two domains, both the marginal distribution and conditional distribution are simultaneously reduced to obtain a more accurate target classifier. Moreover, the predicted results for target data are utilized as pseudo labels to further improve the classification accuracy in multiple iterations. Extensive experiments on public image datasets demonstrate that ELM-MWMD performs better than several existing state-of-the-art domain adaptation methods by computation efficiency and classification accuracy.

Hyperspectral Image Denoising and Classification Using Multi-Scale Weighted EMAPs and Extreme Learning Machine

Recently, extended multi-attribute profiles (EMAPs) have attracted much attention due to its good performance while applied to remote sensing images feature extraction and classification. Since the EMAPs connect multiple attribute features without considering the pixel-based Hyperspectral Image (HSI) classification, homogeneous regions may become unsmooth due to the noise to be introduced. To tackle this problem, we propose the weighted EMAPs (WEMAPs) to reduce the noise and smoothen the homogeneous regions based on weighted mean filter (WMF). Then, we construct multiscale WEMAPs to product multiscale feature in order to extract different spatial structures of the HSI and produce better classification results. Finally, a new joint decision fusion and feature fusion (JDFFF) framework is proposed based on the decision fusion (DF) and the multiscale WEMAPs (MWEMAPs) based on extreme learning machine (ELM) classifier. That is, the classification results from various scales are combined into a final one with ELM to perform the HSI classification. Experiment results show that the proposed algorithm significantly outperforms many state-of-the-art HSI classification algorithms.

Multi-Type Classification Comparison of Mammogram Abnormalities

Cancer is a life-threatening disease which reduces the lifespan of humans. If the disease is treated early, the lifespan can be extended. This paper provides a useful method for detecting the abnormalities in the mammograms. The proposed method uses four phases such as pre-processing, segmentation, feature extraction and classification. In the pre-processing phase, median filter is utilized to enhance the quality of an image. The pre-processed image is then segmented by fuzzy C means (FCM). Three different features such as Gaussian–Hermite moments (GHM), Jacobi moments and pseudo Zernike moments (PZM) are extracted from the segmented image. Finally, extreme learning machine (ELM) classifier identifies the normal, malignant and benign kinds of cancer. This method is compared with four different classifiers. The proposed method is tested on mammographic image analysis society (MIAS) dataset and the performance is evaluated against several analogous approaches in terms of accuracy, sensitivity and specificity. The proposed approach substantially provides the best result.

Classification of neovascularization on retinal images using extreme learning machine

AbstractProliferative diabetic retinopathy is the advanced stage of diabetic retinopathy (DR) resulting in the growth of abnormal vessels on the retinal surface termed as neovascularization. This article primarily deals with the timely detection and classification of retinal images into healthy, neovascularization on optic disc, and elsewhere using an extreme learning machine (ELM) classifier. Initially, a binary mask is employed to enhance the foreground pixels. The resultant image is processed for the extraction of a retinal vessel map using Hessian‐based Frangi filter. For the classification of retinal images, an optimal feature set of 14 features including seven‐moment invariant‐based features are extracted from the vascular map using sequential recursive feature elimination algorithm. Further, the training of the ELM classifier is carried out using the K‐fold cross‐validation technique to improve the performance of the classifier. The proposed method achieves an average accuracy of 98% and an average error rate of less than 0.005 when tested on different globally accessible datasets. Apart from the different statistical results like sensitivity, specificity, and area under curve are estimated as 98.5%, 100%, and 94.2%, respectively, for validating the algorithm. Comparison results reveal that the ELM classifier outperforms the SVM classifier in terms of accuracy and error rate.

Surface Electromyography-Based Muscle Fatigue Analysis Using Binary and Weighted Visibility Graph Features

Surface electromyography (sEMG) is a non-invasive technique to assess the electrical activity of contracting skeletal muscles. sEMG-based muscle fatigue detection plays a key role in sports medicine, ergonomics and rehabilitation. These signals are random, multicomponent, nonlinear and the degree of fluctuations is higher in dynamic contractions. Hence, the extraction of reliable biomarkers remains a challenging task. In this work, an attempt has been made to differentiate non-fatigue, and fatigue conditions using nonlinear techniques, namely, binary and weighted Visibility Graph (VG) features. For this, signals are recorded from the biceps brachii muscle of 52 healthy adult volunteers. These signals are preprocessed, and the contractions associated with the non-fatigue and fatigue conditions are segmented. The graph transformation is performed, and first-order and second-order statistics, along with entropy measures, are extracted from the degree distribution. Parametric and non-parametric machine learning methods are applied for the classification. The results show that the proposed VG approach is able to capture the fluctuations of the signals in non-fatigue and fatigue conditions. Further, all extracted features exhibit a significant difference with [Formula: see text] [Formula: see text]. Maximum accuracy of 89.1% is achieved with information gain selected features and extreme learning machines classifier. Additionally, weighted VG features perform better than the binary version with a difference in the accuracy of 5%. It appears that the proposed approach could be used in real-time implementation for the monitoring of muscle fatigue conditions.

On the implementation of a novel data-intelligence model based on extreme learning machine optimized by bat algorithm for estimating daily chlorophyll-a concentration: Case studies of river and lake in USA

Chlorophyll-a is one of the main indicators for water quality (WQ) analysis in environmental monitoring of aquatic ecosystems. WQ degradation is mostly a result of the increase of the concentration of chlorophyll-a in a waterbody, however, proper estimation of daily chlorophyll-a concentration is a complex problem. In this study, the classic extreme learning machine (ELM), group method of data handling (GMDH), random forest (RF), classification and regression tree (CART), and a novel integrated Bat-ELM model (with the bat optimization algorithm) were developed and applied to predict daily chlorophyll-a (Chl-a) concentration in river and lake ecosystems. Input parameters including turbidity (TU), pH, specific conductance (SC), water temperature (TE), and periodicity were applied as the influential elements for estimating daily Chl-a concentration for two different USGS stations. General results based on RMSE values indicated that the Bat-ELM as the most robust model improved the performance of standard ELM, GMDH, RF, and CART models during the testing procedure by 20.7%, 23.9%, 18.3%, and 27.4% in USGS no. 05543010 and 13.8%, 16.8%, 17.5%, and 52.0% in USGS no. 09014050 in terms of the 7th input-combination, respectively. Moreover, the results revealed that periodicity is the most effective input parameter that considered as the last scenario (input combination) on daily chlorophyll-a (Chl-a) concentration.

Wood Defect Detection Based on Depth Extreme Learning Machine

The deep learning feature extraction method and extreme learning machine (ELM) classification method are combined to establish a depth extreme learning machine model for wood image defect detection. The convolution neural network (CNN) algorithm alone tends to provide inaccurate defect locations, incomplete defect contour and boundary information, and inaccurate recognition of defect types. The nonsubsampled shearlet transform (NSST) is used here to preprocess the wood images, which reduces the complexity and computation of the image processing. CNN is then applied to manage the deep algorithm design of the wood images. The simple linear iterative clustering algorithm is used to improve the initial model; the obtained image features are used as ELM classification inputs. ELM has faster training speed and stronger generalization ability than other similar neural networks, but the random selection of input weights and thresholds degrades the classification accuracy. A genetic algorithm is used here to optimize the initial parameters of the ELM to stabilize the network classification performance. The depth extreme learning machine can extract high-level abstract information from the data, does not require iterative adjustment of the network weights, has high calculation efficiency, and allows CNN to effectively extract the wood defect contour. The distributed input data feature is automatically expressed in layer form by deep learning pre-training. The wood defect recognition accuracy reached 96.72% in a test time of only 187 ms.

Epileptic Seizure Detection using Simulated Annealing based Optimal Feature Subset Selection with Kernel Extreme Learning Machine Classification Model

Epileptic Seizure Detection using Simulated Annealing based Optimal Feature Subset Selection with Kernel Extreme Learning Machine Classification Model

Adaptive online sequential extreme learning machine for dynamic modeling

Extreme learning machine (ELM) is an emerging machine learning algorithm for training single-hidden-layer feedforward networks (SLFNs). The salient features of ELM are that its hidden layer parameters can be generated randomly, and only the corresponding output weights are determined analytically through the least-square manner, so it is easier to be implemented with faster learning speed and better generalization performance. As the online version of ELM, online sequential ELM (OS-ELM) can deal with the sequentially coming data one by one or chunk by chunk with fixed or varying chunk size. However, OS-ELM cannot function well in dealing with dynamic modeling problems due to the data saturation problem. In order to tackle this issue, in this paper, we propose a novel OS-ELM, named adaptive OS-ELM (AOS-ELM), for enhancing the generalization performance and dynamic tracking capability of OS-ELM for modeling problems in nonstationary environments. The proposed AOS-ELM can efficiently reduce the negative effects of the data saturation problem, in which approximate linear dependence (ALD) and a modified hybrid forgetting mechanism (HFM) are adopted to filter the useless new data and alleviate the impacts of the outdated data, respectively. The performance of AOS-ELM is verified using selected benchmark datasets and a real-world application, i.e., device-free localization (DFL), by comparing it with classic ELM, OS-ELM, FOS-ELM, and DU-OS-ELM. Experimental results demonstrate that AOS-ELM can achieve better performance.