Kernel-based Support Vector Machine Research Articles

RETRACTED ARTICLE: Gender classification from face images by mixing the classifier outcome of prime, distinct descriptors

Since the last decade, the area of recognizing gender of a person from an image of his/her face has been playing an important role in the research field. A automatic gender recognition is an important concept, essential for many fields like forensic science and automatic payment system. However, it is very onerous due to high variability factors such as illumination, expression, pose, age, scales, camera quality and occlusion. Humans can easily recognize the difference between genders, but it is a critical task for computer. To overcome this issue, many experimental results have been explained in the existing literature as per the advancement of machine vision. But, still definite optimal solution could not be found. For practical usage, a novel full approach to gender classification which is mainly based on image intensity variation, shape and texture features is proposed in this work. These multi-attribute features are mixed at different spatial scales or levels. The proposed novel system uses two datasets such as Facial ExpressIon Set (FEI) dataset and self-built dataset with various facial expressions. In this research, eight local directional pattern algorithms are used for extracting facial edge feature. Local binary pattern is also used for extracting texture feature, whereas intensity as a added feature. Finally, spatial histograms computed from the above features are concatenated to build a gender descriptor. The proposed descriptor efficiently extracts discriminating information from three different levels, including regional, global and directional level. After the extraction of a gender descriptor, effective linear kernel-based support vector machine superior to other classifiers is used to classify the face image as either male or female. The experimental results show that the classification accuracy obtained with the mixture of outcome of multi-scale, multi-block, distinct and prime feature classification is better than having a single-scaled image. It is worth mentioning that the proposed approach is implemented in MATLAB which achieves an accuracy of 99% on the FEI face dataset (200 faces) and 94% on self-built dataset (200 faces).

A Framework for Optimizing the Deployment of Wireless Sensor Networks

In order to tackle the problem of poor network connectivity and sensing coverage in random wireless sensor networks (WSNs) deployment, this paper proposes hierarchical-logic mapping and deployment algorithms, which use disk communication model and radio frequency propagation model to optimize sensing and connectivity coverage. To further accomplish the optimized approach, the study utilizes efficient image processing algorithm for classification of deployment terrain in order to improve deployment and connectivity coverage with required minimum number of nodes and transmission parameters. The Poisson random distribution process is used to simulate deployment while kernel-based support vector machine is used for terrain classification. The proposed approach is developed into an optimization and visualization framework for characterization and prediction of optimal deployment of WSNs. The performance of the framework is compared with actual small-scale stochastic deployment and existing simulators such as GloMoSim to demonstrate its optimal performance. Results show a significant difference and commonality. With the approach used in this paper, results show that deployment performance can be improved significantly. The availability of the deployment framework will ameliorate the current and future applications of WSN technology such as in military operations, post disaster recovery response, environmental quality, surveillance, and border security.

A Finite State Machine Fall Detection Using Quadrilateral Shape Features

A video-based fall detection system was presented; which consists of data acquisition, image processing, feature extraction, feature selection, classification and finite state machine. A two-dimensional human posture image was represented by 12 features extracted from the generalisation of a silhouette shape to a quadrilateral. The corresponding feature vectors for three groups of human pose were statistically analysed by using a non-parametric Kruskal Wallis test to assess the different significance level between them. From the statistical test, non-significant features were discarded. Four selected kernel-based Support Vector Machine: linear, quadratics, cubic and Radial Basis Function classifiers were trained to classify three human posture groups. Among four classifiers, the last one performed the best in terms of performance matric on testing set. The classifier outperformed others with high achievement ofaverage sensitivity, precision and F-score of 99.19%, 99.25% and 99.22%, respectively. Such pose classification model output was further used in a simple finite state machine to trigger the falling event alarms. The fall detection system was tested on different fall video sets and able to detect the presence offalling events in a frame sequence of videos with accuracy of 97.32% and low computional time.

An efficient feature optimization for wetland mapping by synergistic use of SAR intensity, interferometry, and polarimetry data

Wetlands are home to a great variety of flora and fauna species and provide several unique environmental services. Knowledge of wetland species distribution is critical for sustainable management and resource assessment. In this study, multi-temporal single- and full-polarized RADARSAT-2 and single-polarized TerraSAR-X data were applied to characterize the wetland extent of a test site located in the north east of Newfoundland and Labrador, Canada. The accuracy and information content of wetland maps using remote sensing data depend on several factors, such as the type of data, input features, classification algorithms, and ecological characteristics of wetland classes. Most previous wetland studies examined the efficiency of one or two feature types, including intensity and polarimetry. Fewer investigations have examined the potential of interferometric coherence for wetland mapping. Thus, we evaluated the efficiency of using multiple feature types, including intensity, interferometric coherence, and polarimetric scattering for wetland mapping in multiple classification scenarios. An ensemble classifier, namely Random Forest (RF), and a kernel-based Support Vector Machine (SVM) were also used to determine the effect of the classifier. In all classification scenarios, SVM outperformed RF by 1.5–5%. The classification results demonstrated that the intensity features had a higher accuracy relative to coherence and polarimetric features. However, an inclusion of all feature types improved the classification accuracy for both RF and SVM classifiers. We also optimized the type and number of input features using an integration of RF variable importance and Spearman’s rank-order correlation. The results of this analysis found that, of 81 input features, 22 were the most important uncorrelated features for classification. An overall classification accuracy of 85.4% was achieved by incorporating these 22 important uncorrelated features based on the proposed classification framework.

A new expert system based on hybrid colour and structure descriptor and machine learning algorithms for early glaucoma diagnosis

Medical image classification system is widely used by the radiologists to segment the medical images into meaningful regions. Glaucoma is an optic neuropathy defined by characteristic damage to the optic nerve and accompanying visual field deficits. Early diagnosis and treatment are critical to prevent irreversible vision loss and ultimate blindness. Automatic detection of diabetic retinopathy in retinal image is vital as it delivers data about unusual tissues which is essential for planning treatment. Automating, this method is challenging due to the high variabiity in the appearance of tissue among dissimilar patients and in many circumstances, the comparison between abnormal and normal tissue. This paper presents a new methodology and a computerized diagnostic system for diabetic retinopathy. In this article, adaptive histogram equalization is used to convert colour images to gray scale images followed by significant features are selected using hybrid colour and structure descriptor (HCSD). Finally, various classifiers are used for classification of images into normal and glaucomatous classes. The overall classification accuracy of HCSD with Hybrid Radial Basis Kernel based Support vector Machine (HRKSVM) is 97.55%, HCSD with Support vector Machine (SVM) is 94.77% and HCSD with Hybrid Kernel Support Vector Machine (HKSVM) is 95.71%.

Detection of orchard citrus fruits using a monocular machine vision-based method for automatic fruit picking applications

Due to the variable illumination conditions and occlusion produced by neighbouring fruits and other background participants, vision systems are important in accurately and reliably detecting mature citrus in natural orchard environments for automatic fruit picking applications. A robust citrus fruit detection method based on a monocular vision system was proposed. An adaptive enhanced red and green chromatic map was generated from an illumination-compensated image, which was obtained using block-based local homomorphic filtering. Otsu thresholding, morphology operation, marker-controlled watershed transform and convex hull operation methods were then used in combination to locate potential citrus regions from the chromatic map. Local texture information was extracted from the potential regions using local binary patterns and fed to a histogram intersection kernel-based support vector machine to make the final decision. The performance of the proposed method was evaluated on 127 test images captured in two citrus orchards on both sunny and cloudy days. Under strict PASCAL criteria, the recall rate of correctly detected citrus was greater than 0.86, with 13 false detections.

Multiple incipient fault diagnosis in three-phase electrical systems using multivariate statistical signal processing

This paper presents and evaluates a methodology to detect and diagnose single or multiple faults at their earliest stage in electrical systems. The faults affect the gain, the offset and the phase shifting of the output currents. Following the general Fault Detection and Diagnosis process, the methodology is based on data driven approach for modeling the currents in the time domain, pre-processing with the Park transform and univariate statistical feature extraction and analysis.In the case of incipient faults, the Park transformed currents are more sensitive. Therefore we use their Cumulated Sum (CUSUM) (CUSUM mean or CUSUM variance) for the fault detection. Within the incipient fault ranges (≤10%) and a threshold set to have zero false alarm rate, intensive simulations show that these features successfully detect the fault(s) with a probability of miss detection around 5%.The classification of the seven fault classes that have been identified (3 single and 4 multiple) is successfully done with Linear Discriminant Analysis and Support Vector Machines (SVM) when data is linearly separable or kernel-based SVM when data is non linearly separable. The simulation results show that the misclassification errors are lower than 3%.For the fault estimation, the slope of the CUSUM decision has been identified as a relevant feature. For the different faults (single or multiple), from the evolution of the slope along with the fault severity, an analytical model has been derived. The inversion of this model allows an accurate estimation of the fault level.

Primary User Boundary Detection in Cognitive Radio Networks: Estimated Secondary User Locations and Impact of Malicious Secondary Users

Cognitive radio is an effective technology to improve spectrum utilization through spectrum sharing among primary users (PUs) and secondary users (SUs). Developing accurate and efficient methods to determine the PU coverage is crucial in order to avoid potential interference from SU to PU. In this paper, we propose to use the estimated location of each SU, which is obtained using the self-organizing maps algorithm, to determine the boundary of PU coverage. Furthermore, the PU boundary detection issue considering the presence of malicious SUs is also investigated. Support vector machine (SVM) algorithm, considering both linear and kernel cases, is implemented as the PU boundary detection method. Numerical results show that the boundary detection accuracy based on the estimated network layout approaches the results obtained from an accurate network layout (i.e., each SU with precise location information) and the kernel-based SVM classification results outperform the linear SVM method. In addition, PU boundary detection performance in the presence of malicious SUs is evaluated, in which three different $K$ nearest neighbor based malicious SU detection algorithms are utilized.

The effect of kernel selection on genome wide prediction of discrete traits by Support Vector Machine

This study aimed to compare predictive performance of the linear, radial, polynomial and sigmoid kernel-based Support Vector Machine (SVM) in different scenarios of genetic architecture of a discrete trait (number of quantitative trait loci (QTL) and distribution of QTL effects) as well as heritability level. A 500 centiMorgans genome was simulated consisted of 5 chromosomes, one Morgan each, on which 10,000 bi-allelic single nucleotide polymorphisms (SNP) were distributed. To measures accuracy of prediction, Pearson's correlation between the true and predicted genomic breeding values (rp,t) was estimated. In all of the scenarios studied, increase in the heritability resulted to an increase in the accuracy of prediction. Radial and sigmoid based SVM predictors outperformed polynomial and linear kernels significantly (p < 0.05). The linear and polynomial kernel-based SVMs provided predictions with least accuracy and therefore were not recommended for genome wide prediction of discrete traits. Radial-based SVM had slightly higher predictive performance compared to sigmoid-based SVM, though the differences were non-significant. Therefore, both of them were recommended for predicting genomic breeding values. All of the studied SVM-based predictors had almost the same time and memory requirements.

Linear and Kernel Model Construction Methods for Predicting Drug-Target Interactions in a Chemogenomic Framework.

Identification of drug-target interactions is a crucial process in drug discovery. In this chapter, we present protocols for recent advancements in machine learning methods for predicting drug-target interactions from heterogeneous biological data in a chemogenomic framework, in which prediction is based on the chemical structure data of drug candidate compounds and translated genomic sequence data of target candidate proteins. Most existing methods are based on either linear modeling or kernel modeling. To illustrate linear modeling, we introduce sparsity-induced binary classifiers and sparse canonical correlation analysis. To illustrate kernel modeling, we introduce pairwise kernel-based support vector machines and kernel-based distance learning. Workflows for using these techniques are presented. We also discuss the characteristics of each method and suggest some directions for future research.

Visual and textual features based email spam classification using S-Cuckoo search and hybrid kernel support vector machine

Spam mail classification has been playing a vital role in recent days due to the uncontrollable growth happening in the electronic media. Literature presents several algorithms for email spam classification based on classification methods. In this paper, we propose a spam classification framework using S-Cuckoo and hybrid kernel based support vector machine (HKSVM). At first, the features are extracted from the e-mails based on the text as well as the image. For the textual features, TF-term frequency is used. For the image dependent features, correrlogram and wavelet moment are taken. The hybrid features have then high dimension so the optimum features are identified with the help of hybrid algorithm, called S-Cuckoo search. Then, the classification is done using proposed classifier HKSVM model which is designed based on the hybrid kernel by blending three different kernel functions and then it is used in the SVM classifier. The additional features provided based on image and the modification of SVM classifier provides significant improvement as compared with existing algorithms. The spam classification performance is measured by db1 (combining bare-ling spam and Spam Archive corpus) and db2 (combining lemm-ling spam and Spam Archive corpus). Experimental results show that the proposed spam classification framework has outperformed by having better accuracy of 97.235% when compared with existing approach which is able to achieve only 94.117%.

RETRACTED ARTICLE: A big data classification approach using LDA with an enhanced SVM method for ECG signals in cloud computing

Electrocardiographic (ECG) signals often consist of unwanted noises and speckles. In order to remove the noises, various image processing filters are used in various studies. In this paper, FIR and IIR filters are initially used to remove the linear and nonlinear delay present in the input ECG signal. In addition, filters are used to remove unwanted frequency components from the input ECG signal. Linear Discriminant Analysis (LDA) is used to reduce the features present in the input ECG signal. Support Vector Machines (SVM) is widely used for pattern recognition. However, traditional SVM method does not applicable to compute different characteristics of the features of data sets. In this paper, we use SVM model with a weighted kernel function method to classify more features from the input ECG signal. SVM model with a weighted kernel function method is significantly identifies the Q wave, R wave and S wave in the input ECG signal to classify the heartbeat level such as Left Bundle Branch Block (LBBB), Right Bundle Branch Block (RBBB), Premature Ventricular Contraction (PVC) and Premature Atrial Contractions (PACs). The performance of the proposed Linear Discriminant Analysis (LDA) with enhanced kernel based Support Vector Machine (SVM) method is comparatively analyzed with other machine learning approaches such as Linear Discriminant Analysis (LDA) with multilayer perceptron (MLP), Linear Discriminant Analysis (LDA) with Support Vector Machine (SVM), and Principal Component Analysis (PCA) with Support Vector Machine (SVM). The calculated RMSE, MAPE, MAE, R2 and Q2 for the proposed Linear Discriminant Analysis (LDA) with enhanced kernel based Support Vector Machine (SVM) method is low when compared with other approaches such as LDA with MLP, and PCA with SVM and LDA with SVM. Finally, Sensitivity, Specificity and Mean Square Error (MSE) are calculated to prove the effectiveness of the proposed Linear Discriminant Analysis (LDA) with an enhanced kernel based Support Vector Machine (SVM) method.

Kernel-based support vector machines for automated health status assessment in monitoring sensor data

This paper presents a novel algorithm to assess the health status in monitoring sensor data using a kernel-based support vector machine (SVM) approach. Today, accurate fault prediction is a key issue raised by maintenance. In particular, automatically modelling the normal behaviour from condition monitoring data is probably one of the most challenging problems, specially when there is limited information of real faults. To overcome this difficulty, a data-driven learning framework based on nonparametric density estimation for outlier detection and ν-SVM for normality modelling, with optimal bandwidth selection, is proposed. A health score based on the log-normalisation of the distance to the separating hyperplane is also provided. Experimental results obtained when analysing the propagation of a critical fault over time in a marine diesel engine demonstrate the validity of the algorithm. The predictions of normality models learned were compared to those of the k-nearest neighbours (kNN) method. Low false positive rates on healthy data and improved prediction capacities are achieved.

Fractional DCT and DWT hybridization based efficient feature extraction for gender classification

Fractional discrete Cosine transform (FrDCT), generalisation of discrete Cosine transform (DCT), is a useful mathematical tool that provides flexibility to represent data at any angle α from spatial domain axis. Fusion of discrete wavelet transform (DWT) and FrDCT is used in this paper for gender classification. Initially, 1-level DWT is calculated for 128×128 resized input images. LL sub-band images are then input to FrDCT which transforms the images into u-domain which is at an angle α from spatial domain axis. FrDCT transformed images are used to generate a feature vector based on 8×8 non-overlapping sub-blocks. Dividing an image into sub-blocks helps to remove the local spatial correlation. Thus, proposed technique maintains the computational cost as that in other existing techniques. Rbf kernel based Support Vector Machine (SVM) is used to classify the images on the basis of gender. Images of AT&T, Faces94 and Georgia Tech databases are used to validate the performance of the proposed technique. It is found that proposed technique outperforms as compare to other existing techniques with respect to generalization performance.

Detection of malicious behavior in android apps through API calls and permission uses analysis

SummaryIn recent years, with the prevalence of smartphones, the number of Android malware shows explosive growth. As malicious apps may steal users' sensitive data and even money from mobile and bank accounts, it is important to detect potential malicious behaviors so as to block them. To achieve this goal, we propose a dynamic behavior inspection and analysis framework for malicious behavior detection. A customized Android system is built to record apps' API calls, permission uses, and some other runtime features. We also develop an automated app behavior inspection platform to install and inspect massive samples so as to collect apps' dynamic behavior records. Then these records are exploited to train a string subsequence kernel–based Support Vector Machine (SVM) model, which can be used to classify benign and malicious behaviors offline. To realize online detection, we further extract apps' runtime features including sensitive permission combination uses, sensitive behavior sequences, and user interactions for behavior classification. The classification results can reach an accuracy of 84.9% in offline phase and 99.0% in online phase. Besides, we verify our scheme for identifying malicious apps, and the results show that 71.8% instances of malware samples are identified by running each app for only 18 minutes.

Quadratic-radial-basis-function-kernel for classifying multi-class agricultural datasets with continuous attributes

Classification of agricultural data such as soil data and crop data is significant as it allows the stakeholders to make meaningful decisions for farming. Soil classification aids farmers in deciding the type of crop to be sown for a particular type of soil. Similarly, wheat variety classification assists in selecting the right type of wheat for a particular product. Current methods used for classifying agricultural data are mostly manual. These methods involve agriculture field visits and surveys and are labor-intensive, expensive, and prone to human error. Recently, data mining techniques such as decision trees, k-nearest neighbors (k-NN), support vector machine (SVM), and Naive Bayes (NB) have been used in classification of agricultural data such as soil, crops, and land cover. The resulting classification aid the decision making process of government organizations and agro-industries in the field of agriculture. SVM is a popular approach for data classification. A recent study on SVM highlighted the fact that using multiple kernels instead of a single kernel would lead to better performance because of the greater learning and generalization power. In this work, a hybrid kernel based support vector machine (H-SVM) is proposed for classifying multi-class agricultural datasets having continuous attributes. Genetic algorithm (GA) or gradient descent (GD) methods are utilized to select the SVM parameters C and γ. The proposed kernel is called the quadratic-radial-basis-function kernel (QRK) and it combines both quadratic and radial basis function (RBF) kernels. The proposed classifier has the ability to classify all kinds of multi-class agricultural datasets with continuous features. Rigorous experiments using the proposed method are performed on standard benchmark and real world agriculture datasets. The results reveal a significant performance improvement over state of the art methods such as NB, k-NN, and SVM in terms of performance metrics such as accuracy, sensitivity, specificity, precision, and F-score.

Feature extraction through parallel Probabilistic Principal Component Analysis for heart disease diagnosis

Automatic diagnosis of human diseases are mostly achieved through decision support systems. The performance of these systems is mainly dependent on the selection of the most relevant features. This becomes harder when the dataset contains missing values for the different features. Probabilistic Principal Component Analysis (PPCA) has reputation to deal with the problem of missing values of attributes. This research presents a methodology which uses the results of medical tests as input, extracts a reduced dimensional feature subset and provides diagnosis of heart disease. The proposed methodology extracts high impact features in new projection by using Probabilistic Principal Component Analysis (PPCA). PPCA extracts projection vectors which contribute in highest covariance and these projection vectors are used to reduce feature dimension. The selection of projection vectors is done through Parallel Analysis (PA). The feature subset with the reduced dimension is provided to radial basis function (RBF) kernel based Support Vector Machines (SVM). The RBF based SVM serves the purpose of classification into two categories i.e., Heart Patient (HP) and Normal Subject (NS). The proposed methodology is evaluated through accuracy, specificity and sensitivity over the three datasets of UCI i.e., Cleveland, Switzerland and Hungarian. The statistical results achieved through the proposed technique are presented in comparison to the existing research showing its impact. The proposed technique achieved an accuracy of 82.18%, 85.82% and 91.30% for Cleveland, Hungarian and Switzerland dataset respectively.

Short-term forecasting of saltwater occurrence at La Comté River (French Guiana) using a kernel-based support vector machine

Saltwater intrusion into rivers is a major concern for freshwater exploitation and management in French Guiana (South America). To detect and analyse saltwater occurrence, a permanent station was installed on La Comte River to measure the electrical conductivity C. The objective of the present study was twofold. First, the temporal link between C, sea water level SWL and river discharge Q was explored during the dry seasons from 2009 to 2012 (total measurement duration of ~6 months). A lag of 3 h between C and SWL was evidenced (i.e. the C peaks are delayed by 3 h with high water conditions), as well as the co-occurrence of sea water intrusion with the low Q period. Second, a data-driven approach was set up through a kernel-based support vector machine SVM technique to forecast two events: (1) the forthcoming maximum value of C (for the next 3 h) exceeds 500 µS/cm; (2) C exceeds 500 µS/cm during more than 2 h. One potential drawback of such a data-driven approach is to fail to predict outside the range of calibration: this issue was thoroughly explored by means of an intensive bootstrap-based test exercise. It was showed that SVM has very high degree of predictive capability with accuracy and area under receiver operator curve above 90% in average. We additionally analyse the practical implementation of the SVM model with comparison to alternative popular classification techniques (logistic regression, random forest, linear and quadratic discriminant analysis): the SVM strength is to provide the nonlinear decision boundary without making a priori restrictive assumptions on its shape (like linear or quadratic methods) and without being too sensitive to noisy observations/outliers. Yet this strength can turn to be a weakness unless a careful examination of the shape is done from a physical perspective.

Hyperspectral Image Classification via Kernel Basic Thresholding Classifier

We propose a nonlinear kernel version of recently introduced basic thresholding classifier (BTC) for hyperspectral image (HSI) classification. BTC is a sparsity-based linear classifier, which utilizes inner product for similarity measure based on the fact that the hyperspectral data are linearly separable in the feature space. In practice, the pixels of different classes of a given HSI are not always distinct and may overlap. This is known as the inseparability problem, which reduces the performance of a linear classifier. In order to solve this problem in the feature space, we propose the kernel BTC (KBTC) algorithm that achieves promising performance by utilizing well-known “kernel trick.” To increase the classification accuracy further, we present spatial-spectral KBTC that incorporates spatial information using weighted least squares filter with a guidance image. Furthermore, we apply $L_{0}$ smoothing technique on the guidance image, which provides additional improvements. Experimental results on publicly available HSI data sets showed that the proposal and its spatial extension yield better classification results as compared with linear similarity-based BTC, nonlinear kernel-based support vector machines, kernel orthogonal matching pursuit, recently proposed logistic regression via splitting and augmented Lagrangian, and their spatial extensions.

Saliency-based Pedestrian Detection in Far Infrared Images

Night time pedestrian detection is more and more important in advanced driver assistant systems (ADAS). Traditional pedestrian detection algorithms in far infrared (FIR) images lack accuracy and have long processing times. Focusing on this issue, in this paper, a visual saliency-based pedestrian detection algorithm is proposed. First, areas that contain suspected pedestrians are detected using a fusion saliency-based method. Then, the sub-image of the suspected pedestrian is used as an input to a histogram of local intensity difference feature and cross kernel-based support vector machine classifier to make a final determination. Experiments performed using a real FIR road image data set demonstrated that the proposed fusion saliency-based region of interest (ROI) detection method has the largest pedestrian inclusion rate and the smallest ROI proportion compared with three other methods. Besides, compared with existing state-of-the-art pedestrian detection algorithms, the proposed method demonstrates a much higher pedestrian detection rate with a comparably short processing time.