Multi-class Support Vector Machine Classifier Research Articles

Kernel-Based Constrained Energy Minimization for Hyperspectral Mixed Pixel Classification

One fundamental task of hyperspectral imaging is spectral unmixing. In this case, the conventional pure pixel-based hyperspectral image classification (HSIC) may not work effectively for mixed pixels. This article presents a kernel-based approach to hyperspectral mixed pixel classification (HMPC) which includes two nonlinear mixed pixel classifiers, kernel constrained energy minimization (KCEM) and kernel linearly constrained minimum variance (KLCMV) to replace the widely used pure pixel-based support vector machine (SVM) classifier. Interestingly, what the binary-class and multiclass SVM classifiers are to pure pixel-based HSIC can be similarly derived for what a single-class KCEM detector and a multiclass KLCMV detector are to HMPC. In particular, the commonly used discrete classification map-based hard classification measures, average accuracy (AA) and overall accuracy (OA) for performance evaluation can be further generalized to real-valued mixed class abundance fractional map-based soft classification measures via 3-D receiver operating characteristic (3-D ROC) analysis-derived detection measures. Extensive experiments are conducted to demonstrate the utility of HMPC where KCEM/KLCMV not only significantly improve the classification performance of CEM/LCMV-based classifiers but also outperform many existing spectral-spatial classification methods.

NCodR: A multi-class support vector machine classification to distinguish non-coding RNAs in Viridiplantae

Non-coding RNAs (ncRNAs) are major players in the regulation of gene expression. This study analyses seven classes of ncRNAs in plants using sequence and secondary structure-based RNA folding measures. We observe distinct regions in the distribution of AU content along with overlapping regions for different ncRNA classes. Additionally, we find similar averages for minimum folding energy index across various ncRNAs classes except for pre-miRNAs and lncRNAs. Various RNA folding measures show similar trends among the different ncRNA classes except for pre-miRNAs and lncRNAs. We observe different k-mer repeat signatures of length three among various ncRNA classes. However, in pre-miRs and lncRNAs, a diffuse pattern of k-mers is observed. Using these attributes, we train eight different classifiers to discriminate various ncRNA classes in plants. Support vector machines employing radial basis function show the highest accuracy (average F1 of ~96%) in discriminating ncRNAs, and the classifier is implemented as a web server, NCodR.

Respiratory Effort Signal Based Sleep Apnea Detection System Using Improved Random Forest Classifier

This work aims to develop a minimally obtrusive simple system for the automatic detection and classification of sleep apnea (SA) events using the single-channel abdomen respiratory effort signal (Abd RES) taken from the ISRUC sleep database. The discrete wavelet transform energy, time domain respiratory rate variability and the Shannon entropy features are extracted from the Abd RES to build a classifier system. The performance of linear multi-class support vector machine (SVM) and random forest (RF) classifiers is compared for the application. Furthermore, an improved random forest (IRF) classifier is proposed, which is a combination of data re-sampling, RF algorithm and threshold moving method. It is designed to deal with the imbalance in the input dataset for better classification results. Various systems are designed using individual feature sets, but classification accuracy increases when feature sets are combined. A maximum accuracy of 89.66% is achieved with the IRF based system using reduced feature set obtained by principal component analysis. Thus, the proposed method can provide an effective, economical and accessible screening tool for SA.

Fused-Deep-Features Based Grape Leaf Disease Diagnosis

Rapid and accurate grape leaf disease diagnosis is of great significance to its yield and quality of grape. In this paper, aiming at the identification of grape leaf diseases, a fast and accurate detection method based on fused deep features, extracted from a convolutional neural network (CNN), plus a support vector machine (SVM) is proposed. In the research, based on an open dataset, three types of state-of-the-art CNN networks, three kinds of deep feature fusion methods, seven species of deep feature layers, and a multi-class SVM classifier were studied. Firstly, images were resized to meet the input requirements of the CNN network; then, the deep features of the input images were extracted via the specific deep feature layer of the CNN network. Two kinds of deep features from different networks were then fused using different fusion methods to increase the effective classification feature information. Finally, a multi-class SVM classifier was trained with the fused deep features. The experimental results on the open dataset show that the fused deep features with any kind of fusion method can obtain a better classification performance than using a single type of deep feature. The direct concatenation of the Fc1000 deep feature extracted from ResNet50 and ResNet101 can achieve the best classification result compared with the other two fusion methods, and its F1 score is 99.81%. Furthermore, the SVM classifier trained using the proposed method can achieve a classification performance comparable to that of using the CNN model directly, but the training time is less than 1 s, which has an advantage over spending tens of minutes training a CNN model. The experimental results indicate that the method proposed in this paper can achieve fast and accurate identification of grape leaf diseases and meet the needs of actual agricultural production.

Adaptive Extraction of Oil Painting Texture Features Based on Reaction Diffusion Equation

The oil painting retrieval technology based on the reaction diffusion equation has attracted widespread attention in the fields of oil painting processing and pattern recognition. The description and extraction of oil painting information and the classification method of oil paintings are two important processes in content-based oil painting retrieval. Inspired by the restoration and decomposition functional model of equal oil painting, we propose a reaction diffusion equation model. The new model contains two reaction diffusion equations with different principal parts. One principal part is total variation diffusion, which is used to remove noise. The other main part is thermal diffusion, which is used to modify the source term of the denoising reaction-diffusion equation to achieve the effect of protecting the texture of the oil painting. The interaction of the two reaction-diffusion equations finally achieves denoising while maintaining the boundaries and textures. Under the framework of the above reaction diffusion equation model, we introduce Laplace flow to replace the original total variation flow, so that the new denoising reaction diffusion equation combines the isotropic diffusion and total variation flow of the thermal reaction diffusion equation to achieve the effect of adaptive theoretical research. Using regularization methods and methods, we, respectively, get the well-posedness of the two model solutions, which provides the necessary preparation for numerical calculations. Based on the statistical theory and classification principles of support vector machines, combined with the characteristics of oil painting classification, the research and analysis are carried out from the three important aspects of kernel function, training algorithm, and multiclass classifier algorithm that affect the classification effect and speed. Numerical experiments show that the given filter model has a better processing effect on images with different types and different degrees of noise pollution. On this basis, an oil painting classification system based on texture features is designed, combined with an improved gray-level cooccurrence matrix algorithm and a multiclass support vector machine classification model, to extract, train, and classify oil paintings. Experiments with three types of oil paintings prove that the system can achieve a good oil painting classification effect. Different from the original model, the new model is based on the framework of reaction-diffusion equations. In addition, the new model has good effects in removing step effects, maintaining boundaries and denoising, especially in maintaining texture.

A novel hybrid face recognition framework based on a low-resolution camera for biometric applications

In research work, human face recognition is an essential biometric symbol persistently continued so far due to its different levels of applications in society. Since the appearance of the human faces can have many variations due to issues like the effect of illumination, expression and face pose. These differences are correlated with one another, which results in a helpless ability to recognize a particular person's face. The motivation behind our work in this paper is to give a new framework for face recognition based on frequency analysis that contributes to solving the distinguishing proof issues with enormous varieties of boundaries like the effect of illumination, expression, and face pose. Here three algorithms combined for provable results: i) Difference of Gaussian filtered discrete wavelet transform (DDWT) for feature extraction; ii) Log Gabor (LG) filter for feature extraction; and iv) Multiclass support vector machine classifier, where feature coefficients of DDWT and LG filter are fused for classification and parameters evaluation. The evaluation of our experiment is carried out on a large database consisting of 15 persons of each 200-face image which are captured using a 5-megapixel low-resolution web camera and yielding satisfactory results on various parameters compared to existing methods.

Meningioma DNA Methylation Grouping Reveals Biologic Drivers and Therapeutic Vulnerabilities

Meningiomas arising from the meningothelial lining of the central nervous system are the most common primary intracranial tumors, and a significant cause of neurologic morbidity and mortality. There are no effective medical therapies for meningioma patients, and new treatments have been encumbered by limited understanding of meningioma biology. The aim of this study was to understand the biology underlying meningioma DNA methylation groups, and identify novel therapeutic strategies to treat meningioma patients.Illumina 850K DNA methylation profiling was performed on a discovery cohort of 200 meningiomas (median follow-up 6.3 years) and an external validation cohort of 365 meningiomas (median follow-up 5.3 years) from patients who were treated at 2 independent, international institutions from 1991 to 2019 (69% WHO grade 1, 25% WHO grade 2, 6% WHO grade 3). RNA sequencing was performed on the discovery cohort, and single-cell RNA sequencing was performed on 8 meningiomas to define the cellular architecture of meningiomas across DNA methylation groups. Targeted DNA sequencing (n = 65), H3K27ac ChIP sequencing (n = 25), whole exome sequencing (n = 25), and immunohistochemistry (n = 87) were used to validate novel biologic mechanisms underlying DNA methylation groups.Unsupervised hierarchical clustering of DNA methylation profiles revealed 3 DNA methylation groups of meningioma with distinct clinical outcomes in the discovery cohort, which was validated in the external cohort using a multi-class support vector machine classifier. Merlin-intact meningiomas (34%) had the best outcomes and were distinguished by a novel tumor suppressor function of NF2/Merlin regulating glucocorticoid signaling and apoptosis. Immune-enriched meningiomas (38%) had intermediate outcomes and were distinguished by immune cell infiltration, HLA expression, and lymphatic vessels. Hypermitotic meningiomas (28%) had the worst outcomes and are distinguished by convergent genetic mechanisms misactivating the cell cycle, including CDKN2A/B loss, FOXM1 overexpression, and USF1 gain. Consistently, we found cell cycle inhibitors blocked Immune-enriched and Hypermitotic meningioma growth in cell culture, organoids, xenografts, and patients.Here we integrate genetic, epigenetic, transcriptomic, biochemical, and single-cell approaches to find that meningiomas are comprised of 3 DNA methylation groups with distinct clinical outcomes and biological drivers. We validate our results using mechanistic and functional studies in cells, organoids, xenografts, and patients to elucidate novel mechanisms and identify therapeutic vulnerabilities underlying the most common primary intracranial tumor. Our results establish a framework for understanding meningioma biology, and provide preclinical rationale for new meningioma treatments.A. Choudhury: None. S. Magill: None. C.D. Eaton: None. B.C. Prager: None. W.C. Chen: None. K. Seo: None. C. Lucas: None. J. Villanueva-Meyer: None. H. Vasudevan: Research Grant; Children's Tumor Foundation. Patent/License Fees/Copyright; Genentech, Eli Lilly. S.J. Liu: None. M.A. Cady: None. M.Y. Zhang: None. S.E. Braunstein: Advisory Board; Radiation Oncology Questions, LLC.N.A. Oberheim: None. A. Perry: None. D.A. Solomon: None. J. Costello: None. M.W. McDermott: None. J.N. Rich: None. D. Raleigh: None.

A Leak Zone Location Approach in Water Distribution Networks Combining Data-Driven and Model-Based Methods

Model-based and data-driven methods are commonly used in leak location strategies in water distribution networks. This paper formulates a hybrid methodology in two stages that complements the advantages and disadvantages of data-driven and model-based strategies. In the first stage, a support vector machine multiclass classifier is used to reduce the search space for the leak location task. In the second stage, leak location task is formulated as an inverse problem, and solved using a variation of the differential evolution algorithm called topological differential evolution. The robustness of the method is tested considering measurement and varying demand uncertainty conditions ranging from 5 to 15% of node nominal demands. The performance of the hybrid method is compared to the support vector machine classifier and topological differential evolution approaches as standalone methods of leak location. The hybrid proposal shows higher performance in terms of location accuracy, zone size, and computational load.

Feature Level Combination for Face Recognition Based on Convolutional Neural Networks

Face detection and recognition systems have recently achieved encouraging results using deep learning, especially Convolutional Neural Network (CNN). Face Recognition (FR) systems have many challenges in unconstrained environments that decrease the accuracy; for overcoming these challenges, a deep learning-based features combination has been proposed. The scheme performs feature-level combination by applying two pre-trained GoogLeNet and VggNet-16 models as deep feature extractors. First, faces are detected and aligned using the Multi-Task Convolutional Neural Networks (MTCNN) face detector. The deep features are extracted from a face image using each individually pre-trained CNN. Second, features obtained from GoogLeNet and VggNet-16 models are combined using the serial-feature combination method. Finally, a classification task is performed using a multiclass Support Vector Machine (SVM) classifier. Experiments on the following datasets: VggFace2, LFW, Essex, and ORL, indicate the efficacy of the proposed system as the combination of the two pre-trained CNN models improves performance. The combination strategy, in particular, yields an accuracy of 95.33% to 99.29% on all datasets. The proposed system was compared to existing models in terms of the LFW, and ORL datasets, the findings showed that the proposed system outperformed most current models in terms of accuracy.

Fostering Emotion Recognition in Children with Autism Spectrum Disorder

Facial expressions are of utmost importance in social interactions, allowing communicative prompts for a speaking turn and feedback. Nevertheless, not all have the ability to express themselves socially and emotionally in verbal and non-verbal communication. In particular, individuals with Autism Spectrum Disorder (ASD) are characterized by impairments in social communication, repetitive patterns of behaviour, and restricted activities or interests. In the literature, the use of robotic tools is reported to promote social interaction with children with ASD. The main goal of this work is to develop a system capable of automatic detecting emotions through facial expressions and interfacing them with a robotic platform (Zeno R50 Robokind® robotic platform, named ZECA) in order to allow social interaction with children with ASD. ZECA was used as a mediator in social communication activities. The experimental setup and methodology for a real-time facial expression (happiness, sadness, anger, surprise, fear, and neutral) recognition system was based on the Intel® RealSense™ 3D sensor and on facial features extraction and multiclass Support Vector Machine classifier. The results obtained allowed to infer that the proposed system is adequate in support sessions with children with ASD, giving a strong indication that it may be used in fostering emotion recognition and imitation skills.

Fused Deep Features-Based Grape Varieties Identification Using Support Vector Machine

Proper identification of different grape varieties by smart machinery is of great importance to modern agriculture production. In this paper, a fast and accurate identification method based on Canonical Correlation Analysis (CCA), which can fuse different deep features extracted from Convolutional Neural Network (CNN), plus Support Vector Machine (SVM) is proposed. In this research, based on an open dataset, three types of state-of-the-art CNNs, seven species of deep features, and a multi-class SVM classifier were studied. First, the images were resized to meet the input requirements of a CNN. Then, the deep features of the input images were extracted by a specific deep features layer of the CNN. Next, two kinds of deep features from different networks were fused by CCA to increase the effective classification feature information. Finally, a multi-class SVM classifier was trained with the fused features. When applied to an open dataset, the model outcome shows that the fused deep features with any combination can obtain better identification performance than by using a single type of deep feature. The fusion of fc6 (in AlexNet network) and Fc1000 (in ResNet50 network) deep features obtained the best identification performance. The average F1 Score of 96.9% was 8.7% higher compared to the best performance of a single deep feature, i.e., Fc1000 of ResNet101, which was 88.2%. Furthermore, the F1 Score of the proposed method is 2.7% higher than the best performance obtained by using a CNN directly. The experimental results show that the method proposed in this paper can achieve fast and accurate identification of grape varieties. Based on the proposed algorithm, the smart machinery in agriculture can take more targeted measures based on the different characteristics of different grape varieties for further improvement of the yield and quality of grape production.

DEVELOPMENT OF AUTOMATED CERAMIC TILES SURFACE DEFECT DETECTION AND CLASSIFICATION SYSTEM

This research presented the development of an automated system for ceramic tiles surface defect detection and classification. The production process of ceramic tiles is very fast through the use of automated system except the inspection process that is manually carried out. The fast rate of production and numerous amounts to be produced make it difficult to manually inspect the tiles defects. Currently many literatures have proposed various automated systems for detecting and classifying defects on ceramic tiles. In this research different defected and non-defected images of ceramic tiles were taking at the firing unit in a Ceramic Company with a Nikon D40 camera. A statistical method called Rotation Invariant Measure of Local Variance (RIMLV) operator was used for detection of the defects while morphological operator was used to fill and smooth detected regions. Then, the detected defects are labelled to extract the corresponding features vectors using Fourier descriptors. To categorize the defect, multi-class support vector machine classifier (SVM) was used. The proposed system recorded an accuracy of 98 percent for classification and 0.094939 seconds for classification using one-against all SVM classifier.

Automated Computer Aided Diagnosis Using Altered Multi-Phase Level Sets in Application to Categorize the Breast Cancer Biopsy Images

As macroscopic Breast Cancer (BC) imaging diagnostics do not provide satisfied data, microscopic image analysis is used to get optimum cell-nuclei segmentation for image categorization. The paper presents a Novel Hybrid Segmentation Method which uses the energy detail coefficients attained from Non-Subsampled Contourlet Transform for cell-nuclei segmentation of BC histopathological images. NHSM solves the local inhomogeneities problem while preserving all necessary details. The system evaluation is done using training dataset of 98 BC histopathological images and testing dataset of 26 images. Total of 25 segmented features was extracted for classification using Decision Trees, k-Nearest Neighbor, multilayer perceptron, Multiclass-Support Vector Machine (M-SVM). The proposed NHSM segmentation results were undergone through both subjective and objective performance evaluation. Result evaluation proved that NHSM along with Multi Class-Support Vector Machine classifier provides highly precise and accurate classified result for BC tissue images compared to state-of-the-art segmentation methods. Also, the edge detection model initialization is highly robust and insensitive.

AN EXTENSION OF THE OUTLIER MAP FOR VISUALIZING THE CLASSIFICATION RESULTS OF THE MULTI-CLASS SUPPORT VECTOR MACHINE

The main objective of this study is a graphical display of the results of the high (as well as the low) dimensional multi-class support vector machine classification. Additionally, we will visually be able to detect the outliers and misclassified observations by using this graphical tool. The “outlier map” as a successful graphical outlier detection tool of robust statistics is extended in this paper. In fact, this is a bilateral extension concerning the misclassified and outlying observations recognition. The most important feature of this extension is creating two types of discriminative boundaries to segregate the data and detect the outlying observations. For this purpose, we employed the simple but efficient concept of the “confidence interval”, which is computed for the mean of decision function of support vector machine and then, “thresholding” technique. After that, the efficiency of the outlier map in terms of the preciseness of the correct outlier identification has been tested by the classification accuracy. Moreover, we deployed the margin width “before” and “after” outlier detection as the other criterion to assess the preciseness of the correct outlier identification. We conducted an empirical study based on the proposed method on the simulated and several well-known real datasets. It shows the effectiveness of our proposed method by increasing the “margin width” and gaining a higher classification accuracy.

Deep neural network based robust computer-aided cataract diagnosis system using fundus retinal images

Cataract is the cloudiness present in the eye lens due to denaturation of active protein cells. Cataract affects the quality-of-life and thereby troubling the daily routine activities. Early diagnosis and treatment may reduce the vision loss and delays the cataract progression. To diagnose large-screen population, the computer-aided cataract diagnosis (CACD) system using fundus retinal (FR) images is required. In this paper, a CACD system using FR images is proposed to achieve better diagnostic accuracy. It is perceived that the performance of existing CACD systems is poor against noisy input FR images. However, the distortion such as noise is unavoidable in input images due to complex processes involved in the image acquisition. Hence, it is required to consider the effect of noise in the design of CACD systems. So the proposed CACD system includes this issue in the design and provides the robust performance. In the presented CACD system, the features are extracted using combined feature extraction (CFE) technique using two independently fine-tuned deep convolutional neural networks. The noise level estimation (NLE)-based classification is adopted in the classification stage. In NLE-based classification, a set of multi-class support vector machine (SVM) classifiers, which are trained independently at noise levels from 0 to 25 are considered. Finally, the features extracted using CFE are then mapped to a specific multi-class SVM classifier based on noise level present in an input FR image. From the experimental results, it is observed that the proposed system exhibits superior performance than existing CACD systems under noisy conditions.

Adaptive Accumulation of Plantar Pressure for Ambulatory Activity Recognition and Pedestrian Identification.

In this paper, we propose a novel method for ambulatory activity recognition and pedestrian identification based on temporally adaptive weighting accumulation-based features extracted from categorical plantar pressure. The method relies on three pressure-related features, which are calculated by accumulating the pressure of the standing foot in each step over three different temporal weighting forms. In addition, we consider a feature reflecting the pressure variation. These four features characterize the standing posture in a step by differently weighting step pressure data over time. We use these features to analyze the standing foot during walking and then recognize ambulatory activities and identify pedestrians based on multilayer multiclass support vector machine classifiers. Experimental results show that the proposed method achieves 97% accuracy for the two tasks when analyzing eight consecutive steps. For faster processing, the method reaches 89.9% and 91.3% accuracy for ambulatory activity recognition and pedestrian identification considering two consecutive steps, respectively, whereas the accuracy drops to 83.3% and 82.3% when considering one step for the respective tasks. Comparative results demonstrated the high performance of the proposed method regarding accuracy and temporal sensitivity.

Optimized Variational Mode Decomposition and Permutation Entropy with Their Application in Feature Extraction of Ship-Radiated Noise.

The complex and changeable marine environment surrounded by a variety of noise, including sounds of marine animals, industrial noise, traffic noise and the noise formed by molecular movement, not only interferes with the normal life of residents near the port, but also exerts a significant influence on feature extraction of ship-radiated noise (S-RN). In this paper, a novel feature extraction technique for S-RN signals based on optimized variational mode decomposition (OVMD), permutation entropy (PE), and normalized Spearman correlation coefficient (NSCC) is proposed. Firstly, with the mode number determined by reverse weighted permutation entropy (RWPE), OVMD decomposes the target signal into a set of intrinsic mode functions (IMFs). The PE of all the IMFs and SCC between each IMF with the raw signal are then calculated, respectively. Subsequently, feature parameters are extracted through the sum of PE weighted by NSCC for the IMFs. Lastly, the obtained feature vectors are input into the support vector machine multi-class classifier (SVM) to discriminate various types of ships. Experimental results indicate that five kinds of S-RN samples can be accurately identified with a recognition rate of 94% by the proposed scheme, which is higher than other previously published methods. Hence, the proposed method is more advantageous in practical applications.

Human Activity Recognition Algorithm in Video Sequences Based on Integration of Magnitude and Orientation Information of Optical Flow

Human activity recognition from video sequences has emerged recently as pivotal research area due to its importance in a large number of applications such as real-time surveillance monitoring, healthcare, smart homes, security, behavior analysis, and many more. However, lots of challenges also exist such as intra-class variations, object occlusion, varying illumination condition, complex background, camera motion, etc. In this work, we introduce a novel feature descriptor based on the integration of magnitude and orientation information of optical flow and histogram of oriented gradients which gives an efficient and robust feature vector for the recognition of human activities for real-world environment. In the proposed approach first we computed magnitude and orientation of the optical flow separately then a local-oriented histogram of magnitude and orientation of motion flow vectors are computed using histogram of oriented gradients followed by linear combination feature fusion strategy. The resultant features are then processed by a multiclass Support Vector Machine (SVM) classifier for activity recognition. The experimental results are performed over different publically available benchmark video datasets such as UT interaction, CASIA, and HMDB51 datasets. The effectiveness of the proposed approach is evaluated in terms of six different performance parameters such as accuracy, precision, recall, specificity, [Formula: see text]-measure, and Matthew’s correlation coefficient (MCC). To show the significance of the proposed method, it is compared with the other state-of-the-art methods. The experimental result shows that the proposed method performs well in comparison to other state-of-the-art methods.

Approach to hand posture recognition based on hand shape features for human–robot interaction

Hand segmentation is the initial step for hand posture recognition. To reduce the effect of variable illumination in hand segmentation step, a new CbCr-I component Gaussian mixture model (GMM) is proposed to detect the skin region. The hand region is selected as a region of interest from the image using the skin detection technique based on the presented CbCr-I component GMM and a new adaptive threshold. A new hand shape distribution feature described in polar coordinates is proposed to extract hand contour features to solve the false recognition problem in some shape-based methods and effectively recognize the hand posture in cases when different hand postures have the same number of outstretched fingers. A multiclass support vector machine classifier is utilized to recognize the hand posture. Experiments were carried out on our data set to verify the feasibility of the proposed method. The results showed the effectiveness of the proposed approach compared with other methods.

A new feature extraction approach for script invariant handwritten numeral recognition

AbstractHandwritten numeral recognition is a challenging research problem because of the enormous varieties of styles in which human beings write the numerals. Several researchers have tried to find solutions to this problem with exceptional recognition accuracies. However, most of these solutions have been dedicated to single script numerals. Such methods are inappropriate for multi‐lingual nations such as India where a large number of scripts are used. Keeping this issue in mind, a new feature descriptor named symbolization of binary images (SBI) is introduced here for the recognition of handwritten numerals of different scripts. Effectiveness of SBI is supported with experiments showing its script‐invariant nature. Classification of numerals using a multiclass support vector machine (SVM) classifier yields the recognition accuracies of 98.18, 96.22, 96.52, and 95.53% on datasets of numerals written in four popular scripts of the world: Arabic, Bangla, Devanagari, and Latin, respectively. This scheme has also been extended to the situation when the script used is not known a priori or the numerals written in a document belong to pairs of mixed scripts of {Arabic, Devanagari, Bangla} with Latin producing recognition rates of 92.97, 91.25, and 91.67%, respectively. When all four scripts are mixed, the recognition rate is still 90.98% overall. Encouraging outcomes suggest that the proposed SBI feature descriptor can recognize numerals invariant of the script class.