Parzen Classifier Research Articles

The design for supply chain management of intelligent logistics system using cloud computing and the internet of things

AbstractImage recognition is the key to smart logistics systems. Traditional handwriting feature extraction is difficult to meet the requirements of image recognition. Deep learning is used for image recognition. Firstly, convolutional neural network (CNN) and deep Boltzmann machines under deep learning are introduced. Second, cellular neural networks are used to perform feature recognition and extraction on images. Finally, a Parzen classifier is used to classify the obtained image features. The novelty is that through the structural design and research of the intelligent logistics system, the CNN is combined to construct a management system of supply chain logistics of image recognition and information processing. The experimental results show that the recognition accuracy time of the proposed improved fusion algorithm on the Mixed National Institute of Standards and Technology data set is 198.85 s. When the improved algorithm achieves the same recognition accuracy, it takes 159.65 s. The recognition efficiency of the improved algorithm is 19.71% higher than that of the unimproved algorithm. In addition, when the unimproved algorithm reaches the maximum number of iterations, the error rate is 2.47%. The error rate of the improved algorithm is only 0.74%. This study provides a basis for improving the image recognition accuracy and has certain practical value.

Application of Clustering Techniques on Statistical Features of EEG Signals for Seizure Detection

Epileptic seizure is the most predominant brain disorder that has affected at least 1% of the world's population and higher numbers are affected in the underdeveloped countries. The best marker for diagnosis of seizure is to analyse the EEG signals. The statistical parameters of EEG signals vary when there is an occurrence of seizure. In this paper the statistical parameter interquatile range and Mean absolute deviation are taken as features and are classified using unsupervised learning algorithms (clustering techniques). K-Mean, K-Centers, Parzen Classifier, K-Nearest Neighbour, Gaussian mixture model, and Naive Bayes classifier are put to test to find out the best suited algorithm. Among the methods used K-Nearest Neighbour gives the maximum accuracy of 100%, sensitivity and specificity of 1 for the given data followed by Parzen classifier with 98.02% of accuracy, and has a sensitivity of 0.96, specificity of 1. The Naieve bayes classifier with 96.71% of accuracy is the fastest algorithm with 0.019 ms and hence can be used for real time applications.

Single‐signal entity approach for sung word recognition with artificial neural network and time–frequency audio features

Singing voice recognition is very different from speech recognition or automatic speech recognition because there are distinct differences between speaking and singing voices. The problem is complex because music audio signals with their background instrumental accompaniments are regarded as noise sources that degrade the performance of the recognition system. This study proposes a statistical learning method to recognise words in a vocal audio signal with background music and to classify the region of a singing voice in a polyphonic audio signal. The goal of this study is to solve the problem of recognising words from sung input without using any method to separate instrumental from the background. This study also applies a concept from image recognition by using a spectrogram feature as an image to solve the problem. An audio signal with accompanying music was analysed and transformed into a spectrogram feature. To recognise it, the entire spectrogram feature was sliced, forming a feature vector for a feed-forward neural network classifier. Several classification functions were compared, including K-Nearest Neighbour, Fisher Linear Classifier, Linear Bayes Normal Classifier, Naive Bayes Classifier, Parzen Classifier and Decision Tree. The results show that using a feed-forward neural network can effectively recognise sung words at an accuracy rate of more than 93.0%. In particular, this system can recognise cross-language music data.

Rock physics-based carbonate pore type identification using Parzen classifier

Seismic velocity variation in carbonate rocks is a complicated function of different parameters such as mineral composition, porosity, pore type, saturation, and pore pressure. Among all, pore type is the main factor that affects reservoir permeability heterogeneity and change the velocity–porosity relationship. In this paper, a rock physics-based algorithm is presented to quantitatively identify three dominant pore types in a carbonate reservoir. The proposed algorithm is applied on data related to three wells drilled in a carbonate reservoir, southwest of Iran. We used the frame flexibility factor (γ), P-wave velocity–porosity and S-wave impedance–porosity trends as inputs of Parzen classifier to identify predominate pore type characterized by velocity-deviation log (VDL) in each depth. The results show that the proposed algorithm has high precision in classifying identified pore types with average accuracy of 76.7% throughout studied oil field.

Automated Diagnosis of Mammogram Images of Breast Cancer Using Discrete Wavelet Transform and Spherical Wavelet Transform Features: A Comparative Study

Mammograms are one of the most widely used techniques for preliminary screening of breast cancers. There is great demand for early detection and diagnosis of breast cancer using mammograms. Texture based feature extraction techniques are widely used for mammographic image analysis. In specific, wavelets are a popular choice for texture analysis of these images. Though discrete wavelets have been used extensively for this purpose, spherical wavelets have rarely been used for Computer-Aided Diagnosis (CAD) of breast cancer using mammograms. In this work, a comparison of the performance between the features of Discrete Wavelet Transform (DWT) and Spherical Wavelet Transform (SWT) based on the classification results of normal, benign and malignant stage was studied. Classification was performed using Linear Discriminant Classifier (LDC), Quadratic Discriminant Classifier (QDC), Nearest Mean Classifier (NMC), Support Vector Machines (SVM) and Parzen Classifier (ParzenC). We have obtained a maximum classification accuracy of 81.73% for DWT and 88.80% for SWT features using SVM classifier.

Vehicle identification by improved stacking via kernel principal component regression

Purpose – Many applications in intelligent transportation demand accurate categorization of vehicles. The purpose of this paper is to propose a working image-based vehicle classification system. The first component vehicle detection is implemented by applying Dalal and Triggs's histograms of oriented gradients features and linear support vector machine (SVM) classifier. The second component vehicle classification, which is the emphasis of this paper, is accomplished by an improved stacked generalization. As an effective ensemble learning strategy, stacked generalization has been proposed to combine multiple models using the concept of a meta-learner. However, it was found that the well-known meta-learning scheme multi-response linear regression (MLR) for stacked generalization performs poorly on the vehicle classification. Design/methodology/approach – A new meta-learner is then proposed based on kernel principal component regression (KPCR). The stacked generalization scheme consists of a heterogeneous classifier ensemble with seven base classifiers, i.e. linear discriminant classifier, fuzzy k-nearest neighbor, logistic regression, Parzen classifier, Gaussian mixture model, multiple layer perceptron and SVM. Findings – Experimental results using more than 2,500 images from four types of vehicles (bus, light truck, car and van) demonstrated the effectiveness of the proposed approach. The improved stacked generalization produced consistently better results when compared to any of the single base classifier used and four other beta learning algorithms, including MLR, majority voting, logistic regression and decision template. Originality/value – With the seven base classifiers, the KPCR-based stacking offers a performance of 96 percent accuracy and 95 percent κ coefficient, thus exhibiting promising potentials for real-world applications.

A comparative study of two data-driven methods in detection of vuggy zones: a case study from a carbonate reservoir, west of Iran

Vuggy porosity is a key pore space type associated with the characterization of any carbonate reservoir. This porosity has significant influence on many parameters in reservoirs such as permeability, oil recovery, pressure drop and production. In this paper, two data-driven methods (i.e., Bayesian and Parzen classifiers) are used to identify vuggy porosity zones from well logs. Four well logs namely gamma ray (GR), neutron porosity (NPHI), bulk density (RHOB) and sonic (DT) from three oil wells drilled in Sarvak Formation, west of Iran, were integrated to examine the capability of the methods. The results show that the Parzen classifier gives better results than Bayesian in terms of classification accuracy. The Parzen technique classifies the given depth with an average accuracy of 77.7 % in single-well analysis and average accuracy of 67.3 % in generalization step.

Zonal Classification by Pattern Recognition Methods: An Example from Asmari Formation (Mansuri Oil Field, South of Iran)

Uncertainty and risk normally permeate our decisions on reservoirs characterization. A key to mitigate the uncertainties and ambiguities in our analysis is to use some methods that are more quantitative in lieu of the existing conventional one. The present study focuses on identifying first three zones of Asmari Formation as well as its cap rock (Mansuri oil field, south of Iran) from Petrophysical Logs (PLs) using three data-driven methods (i.e. Bayesian classifier, KNN classifier, and Parzen classifier). In term of Lithology, Asmari Formation mainly consists of carbonates and sandstones. In order to generate suitable input data, fluid substitution was performed using Gassmann equation. According to the results, KNN classifier is the best classifier for identifying zones of Asmari Formation. Generalization of the proposed method was also examined and the results of this stage were combined with the use of optimistic Ordered Weighted Average (OWA) technique. The employed fusion technique has improved the results of this step up to 7.6%.

Decision support system for breast cancer detection using mammograms

Mammograms are by far one of the most preferred methods of screening for breast cancer. Early detection of breast cancer can improve survival rates to a greater extent. Although the analysis and diagnosis of breast cancer are done by experienced radiologists, there is always the possibility of human error. Interobserver and intraobserver errors occur frequently in the analysis of medical images, given the high variability between every patient. Also, the sensitivity of mammographic screening varies with image quality and expertise of the radiologist. So, there is no golden standard for the screening process. To offset this variability and to standardize the diagnostic procedures, efforts are being made to develop automated techniques for diagnosis and grading of breast cancer images. This article presents a classification pipeline to improve the accuracy of differentiation between normal, benign, and malignant mammograms. Several features based on higher-order spectra, local binary pattern, Laws' texture energy, and discrete wavelet transform were extracted from mammograms. Feature selection techniques based on sequential forward, backward, plus-l-takeaway-r, individual, and branch-and-bound selections using the Mahalanobis distance criterion were used to rank the features and find classification accuracies for combination of several features based on the ranking. Six classifiers were used, namely, decision tree classifier, fisher classifier, linear discriminant classifier, nearest mean classifier, Parzen classifier, and support vector machine classifier. We evaluated our proposed methodology with 300 mammograms obtained from the Digital Database for Screening Mammography and 300 mammograms from the Singapore Anti-Tuberculosis Association CommHealth database. Sensitivity, specificity, and accuracy values were used to compare the performances of the classifiers. Our results show that the decision tree classifier demonstrated an excellent performance compared to other classifiers with classification accuracy, sensitivity, and specificity of 91% for the Digital Database for Screening Mammography database and 96.8% for the Singapore Anti-Tuberculosis Association CommHealth database.

Parzen classification in generalised dissimilarity spaces

The dissimilarity representation for pattern recognition has proved to be a feasible alternative to the traditional feature based approach. It is shown that the application of the Parzen classifier, in the so-called generalised dissimilarity space by feature lines, may improve the recognition performance obtained by other classification rules in the original dissimilarity space and in the generalised dissimilarity space.

Vision-based Classification of Driving Postures by Efficient Feature Extraction and Bayesian Approach

Reports of traffic accidents show that a considerable percentage of the accidents are caused by human factors. Human-centric driver assistance systems, with integrated sensing, processing and networking, aim to find solutions to this problem and other relevant issues. The key technology in such systems is the capability to automatically understand and characterize driver behaviors. In this paper, we propose a novel, efficient feature extraction approach for driving postures from a video camera, which consists of Homomorphic filter, skin-like regions segmentation, canny edge detection, connected regions detection, small connected regions deletion and spatial scale ratio calculation. With features extracted from a driving posture dataset we created at Southeast University (SEU), holdout and cross-validation experiments on driving posture classification are then conducted using Bayes classifier. Compared with a number of commonly used classification methods including naive Bayes classifier, subspace classifier, linear perception classifier and Parzen classifier, the holdout and cross-validation experiments show that the Bayes classifier offers better classification performance than the other four classifiers. Among the four predefined classes, i.e., grasping the steering wheel, operating the shift gear, eating a cake and talking on a cellular phone, the class of talking on a cellular phone is the most difficult to classify. With Bayes classifier, the classification accuracies of talking on a cellular phone are over 90 % in holdout and cross-validation experiments, which shows the effectiveness of the proposed feature extraction method and the importance of Bayes classifier in automatically understanding and characterizing driver behaviors towards human-centric driver assistance systems.

Recognition of driving postures by multiwavelet transform and multilayer perceptron classifier

To develop Human-centric Driver Assistance Systems (HDAS) for automatic understanding and charactering of driver behaviors, an efficient feature extraction of driving postures based on Geronimo–Hardin–Massopust (GHM) multiwavelet transform is proposed, and Multilayer Perceptron (MLP) classifiers with three layers are then exploited in order to recognize four pre-defined classes of driving postures. With features extracted from a driving posture dataset created at Southeast University (SEU), the holdout and cross-validation experiments on driving posture classification are conducted by MLP classifiers, compared with the Intersection Kernel Support Vector Machines (IKSVMs), the k-Nearest Neighbor (kNN) classifier and the Parzen classifier. The experimental results show that feature extraction based on GHM multwavelet transform and MLP classifier, using softmax activation function in the output layer and hyperbolic tangent activation function in the hidden layer, offer the best classification performance compared to IKSVMs, kNN and Parzen classifiers. The experimental results also show that talking on a cellular phone is the most difficult one to classify among four predefined classes, which are 83.01% and 84.04% in the holdout and cross-validation experiments respectively. These results show the effectiveness of the feature extraction approach using GHM multiwavelet transform and MLP classifier in automatically understanding and characterizing driver behaviors towards Human-centric Driver Assistance Systems (HDAS).

Classification of Electroencephalographic Changes in Meditation and Rest: using Correlation Dimension and Wavelet Coefficients

Meditation is a practice of concentrated focus upon the breath in order to still the mind. In this paper we have investigated an algorithm to classify rest and meditation, by processing of electroencephalogram (EEG) signals through the Wavelet and nonlinear methods. For this purpose, two types of EEG time series (before, and during meditation) of 25 healthy women are collected in the meditation clinic in Mashhad. Correlation dimension and Wavelet coefficients at the forth decomposition level of EEG signals in Fz, Cz and Pz are extracted and used as an input of different classifiers. In order to evaluate performance of the classifiers, the classification accuracies and mean square error (MSE) of the classifiers were examined. The results show that the Fisher discriminant and Parzen classifier trained on both composite features obtain higher accuracy than that of the others. The total classification accuracy of the Fisher discriminant and Parzen classifier applying Wavelet coefficients was 85.02% and 84.75%, respectively which is raised to 92.37% in both classifiers using Correlation dimensions.

Investigating the validity of conventional joint set clustering methods

Up to 10 properties of joints can be recorded in the field, yet only two (dip and dip direction) are commonly used to identify joint sets. This paper investigates some of the shortcomings of commonly employed methods for joint set clustering, based on an analysis of synthetic and field data. First, eight synthetic joint sets were generated using a normal distribution of joint orientations. Each joint was defined in terms of four properties (dip, dip direction, infilling material and infilling percentage). A Parzen classifier was used to confirm the importance of using all the joint properties in identifying the joint sets. To investigate the generalization ability of this approach, the analysis was extended to 178 joints measured in the field, with seven properties available for each joint. Joints were clustered based on rose diagrams, stereonets, and K-means clustering methods, yielding three, five, and seven joint sets, respectively. Calculation of the coefficient of variation and principal component analysis (PCA) of joint properties resulted in an improvement in clustering, provided that a large number of joint properties are considered.

E-Nose Vapor Identification Based on Dempster–Shafer Fusion of Multiple Classifiers

Electronic noses (e-noses) are commonly used to monitor air contaminants in space stations and shuttles. Data preprocessing (measurement denoising and feature extraction) and pattern classification are important problems of an e-nose system. In this paper, the application of a wavelet-based denoising method and a Dempster-Shafer (DS) classification fusion method in an e-nose system are proposed. Six transient-state features are extracted from the sensor measurements filtered by the wavelet denoising method and are used to train multiple classifiers such as multilayer perceptrons (MLPs), support vector machines (SVMs), k -nearest neighbors (KNNs), and the Parzen classifier. The DS technique is used at the end to fuse the results of the multiple classifiers to get the final classification. Experimental analysis based on real vapor data shows that the wavelet denoising method can successfully remove both random noise and outliers, and the classification rate can be improved by using classifier fusion.

Improvement of the Parzen classifier in small training sample size situations

In this paper, we discuss the improvement of the generalization ability of Parzen classifiers, in small sample, high-dimensional setting. When the sizes of samples per class are much unequal, the performance of the Parzen classifier is further degraded. Also, in a high-dimensional space, the degradation becomes clear. In order to overcome this problem, we propose to use the Toeplitz estimator and bootstrap samples in designing Parzen classifiers. Experimental results show that these two techniques are very effective means for designing Parzen classifiers, particularly when the sizes of samples per class are much unequal, or when the number of features is large.

An Integrated Framework for Generalized Nearest Prototype Classifier Design

We propose a Generalized Nearest Prototype Classifier (GNPC) as a common framework for a number of classification techniques. Specifically we consider clustering-and-relabeling; Parzen's classifier; radial basis functions (RBF) networks; learning vector quantization (LVQ) type classifiers; and nearest neighbor rules. To classify an unlabeled point x the GNPC combines the degrees of similarity of x to a set of prototypes. Five questions are addressed for these GNPC families: (1) How many prototypes do we need? (2) How are the prototypes found? (3) How are their class labels obtained? (4) How are the similarities defined? (5) How are the similarities and label information combined? The classification performance of a set of GNPCs is illustrated on two benchmark data sets: IRIS and the 2-spirals data. We study the resubstitution error of the GNPC as a function of the number of prototypes. Our conclusions are that: (a) unsupervised selection (or extraction) of prototypes followed by relabeling is inferior to the techniques that use labels to guide them towards prototypes; (b) the edited nearest neighbor rule is a viable option for GNPC design which has not received the attention it deserves.

A Parzen classifier with an improved robustness against deviations between training and test data

A key assumption in statistical pattern recognition, is that the data that is used to design a classifier is sufficiently representative for the underlying recognition problem. In many practical applications this is only partly true, as the probability density functions of the data may vary as a function of time and/or place. This implies, for example, that a classifier that is trained with data from a certain location, does not necessarily perform well when it is applied to data from another location. In this paper we present a classification method that tries to anticipate slight deviations of the (statistical properties of the) training and the test data (i.e. the data to which the classifier is applied in the operational phase). This method is based on a generalised likelihood function for determining the kernel width of the Parzen classifier. For a simplified case, we will prove that the new classifier is indeed more robust against deviating training and test data than the original Parzen classifier. Furthermore, in an experimental evaluation, the new method is found to have a superior behaviour when applied to the problem of lithofacies recognition from wireline logs.

The reduced Parzen classifier

The Parzen density estimate is known to be an effective tool for estimating the Bayes error, given a set of training samples from the class distributions. An algorithm is developed to select a given number of representative samples whose Parzen density estimate closely matches that of the entire sample set. Using this reduced representative set, a piecewise quadratic classifier which provides nearly optimal performance is designed.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>