Back Propagation Neural Network Classifier Research Articles

FPGA-based electrocardiography (ECG) signal analysis system using least-square linear phase finite impulse response (FIR) filter

This paper presents a proposed design for analyzing electrocardiography (ECG) signals. This methodology employs highpass least-square linear phase Finite Impulse Response (FIR) filtering technique to filter out the baseline wander noise embedded in the input ECG signal to the system. Discrete Wavelet Transform (DWT) was utilized as a feature extraction methodology to extract the reduced feature set from the input ECG signal. The design uses back propagation neural network classifier to classify the input ECG signal. The system is implemented on Xilinx 3AN-XC3S700AN Field Programming Gate Array (FPGA) board. A system simulation has been done. The design is compared with some other designs achieving total accuracy of 97.8%, and achieving reduction in utilizing resources on FPGA implementation.

A Wavelet Based Multiscale Weighted Permutation Entropy Method for Sensor Fault Feature Extraction and Identification

Sensor is the core module in signal perception and measurement applications. Due to the harsh external environment, aging, and so forth, sensor easily causes failure and unreliability. In this paper, three kinds of common faults of single sensor, bias, drift, and stuck-at, are investigated. And a fault diagnosis method based on wavelet permutation entropy is proposed. It takes advantage of the multiresolution ability of wavelet and the internal structure complexity measure of permutation entropy to extract fault feature. Multicluster feature selection (MCFS) is used to reduce the dimension of feature vector, and a three-layer back-propagation neural network classifier is designed for fault recognition. The experimental results show that the proposed method can effectively identify the different sensor faults and has good classification and recognition performance.

Investigating machine learning techniques for the detection of autism

Automated autism detection is needed to facilitate urgently required therapy. However, contrary to cancer, autism detection using genetic data has not attracted much attention. In this paper, we investigate autism detection using machine learning techniques. The main goal is to test whether genetic data with machine learning tools can result in an abbreviated and accurate instrument for classification of autism. For this, a system comprising four stages is proposed, where at each stage, we experiment with different feature reduction, classification and combination methods to find if it is possible to detect autism. The experimental results show that our classifier-based system can achieve optimum accuracy of early screening. We achieved optimum accuracy when examined on independent and unseen test data. The optimum performance was mostly achieved using a three-layer back-propagation neural network classifier combined using the feature selection-based combiner. This was achievable only when the data dimensionality was reduced using our proposed feature selection method. The maximum number of features varied for the different chromosomes and ranged between 150 and 500.

A Cooperative and Hybrid Network Intrusion Detection Framework in Cloud Computing Based on Snort and Optimized Back Propagation Neural Network

Cloud computing provides a framework for supporting end users easily attaching powerful services and applications through Internet. To give secure and reliable services in cloud computing environment is an important issue. Providing security requires more than user authentication with passwords or digital certificates and confidentiality in data transmission, because it is vulnerable and prone to network intrusions that affect confidentiality, availability and integrity of Cloud resources and offered services. To detect DoS attack and other network level malicious activities in Cloud, use of only traditional firewall is not an efficient solution. In this paper, we propose a cooperative and hybrid network intrusion detection system (CH-NIDS) to detect network attacks in the Cloud environment by monitoring network traffic, while maintaining performance and service quality. In our NIDS framework, we use Snort as a signature based detection to detect known attacks, while for detecting network anomaly, we use Back-Propagation Neural network (BPN). By applying snort prior to the BPN classifier, BPN has to detect only unknown attacks. So, detection time is reduced. To solve the problem of slow convergence of BPN and being easy to fall into local optimum, we propose to optimize the parameters of it by using an optimization algorithm in order to ensure high detection rate, high accuracy, low false positives and low false negatives with affordable computational cost. In addition, in this framework, the IDSs operate in cooperative way to oppose the DoS and DDoS attacks by sharing alerts stored in central log. In this way, unknown attacks that were detected by any IDS can easily be detected by others IDSs. This also helps to reduce computational cost for detecting intrusions at others IDS, and improve detection rate in overall the Cloud environment.

Fast HEVC Coding Unit Decision Based On BP-Neural Network

High Efficiency Video Coding (HEVC) employs a highly flexible quad-tree coding block partitioning structure characterized by a coding unit (CU), prediction unit (PU) and transform unit (TU). Taking the computational complexity of CU splitting in HEVC into consideration, a novel fast CU selection algorithm based on a back-propagation neural network (BPNN) is proposed in this paper. To find the best prediction mode at the current CU, the BPNN classifier is designed for two kinds of outputs representing “split” and “unsplit”. Thus, whether to split the current CU or not is directly judged by the outputs of the BPNN classifier. With this new scheme, the unnecessary CU splitting process can be skipped in advance. Experiments and results show that the proposed algorithm could greatly reduce the high computational complexity and the average time savings reaches 47.75% in Random Access Main configuration and 41.94% in Low Delay B Main configuration with a BD-bit rate increase of 1.89%.

A Novel Machine Vision System for the Inspection of Micro-Spray Nozzle.

In this study, we present an application of neural network and image processing techniques for detecting the defects of an internal micro-spray nozzle. The defect regions were segmented by Canny edge detection, a randomized algorithm for detecting circles and a circle inspection (CI) algorithm. The gray level co-occurrence matrix (GLCM) was further used to evaluate the texture features of the segmented region. These texture features (contrast, entropy, energy), color features (mean and variance of gray level) and geometric features (distance variance, mean diameter and diameter ratio) were used in the classification procedures. A back-propagation neural network classifier was employed to detect the defects of micro-spray nozzles. The methodology presented herein effectively works for detecting micro-spray nozzle defects to an accuracy of 90.71%.

Role of GLCM Features in Identifying Abnormalities in the Retinal Images

Accurate detection of exudates in the diabetic retinal images is a challenging task. The images can have varying contrast and color characteristics. In this paper authors present the performance comparison of two feature extraction methods namely color intensity features and second order texture features based on GLCM. Authors have proposed and implemented new approach for GLCM feature calculation in which the input image is divided into number smaller blocks and GLCM features are computed on these blocks. The performance of each feature extraction method is evaluated using Back Propagation Neural Network (BPNN) classifier that is classifying the blocks as either abnormal block or normal block. With GLCM features, an accuracy of 76.6% was obtained and with color features an accuracy of 100% was obtained. It was found that color features are better in identifying true positives than GLCM based texture features. However use of GLCM features reduces the occurrence of false positives. Index Terms—Texture features, Hard exudates, GLCM features, Back propagation neural network.

Face Recognition using Two Dimension Fractional Discrete Cosine Domain and BPNN

Face database contains images taken at various instants of the same person. The matching accuracy of spatial features drops significantly both in the presence of noise as well as when the variations in the different instances are high. To attenuate the image variations up to certain extent, image data is transformed from spatial domain to transformed domain. In this paper an effort is made to explore the effect of using fractional order spectrum obtained by the application of 2D FRDCT on the accuracy of face recognition. PCA is used as dimension reduction approach for reducing transformed feature set dimensionality. Reduced feature set is then classified by back propagation neural network classifier. Through the experiments performed on AT&T database it is shown that proposed FRDCT feature set approach gives a recognition accuracy of 94% with BPNN. Comparison is conducted for fractional order feature classification accuracy of AT&T public database with nearest neighbour classification approach. Experimental result shows marked reduction in classification error rate with neural network classification. General Terms Face recognition, Fractional DCT.

Automated inspection of micro-defect recognition system for color filter

This study focused on micro-defect recognition and classification in color filters. First, six types of defects were examined, namely grain, black matrix hole (BMH), indium tin oxide (ITO) defect, missing edge and shape (MES), highlights, and particle. Orthogonal projection was applied to locate each pixel in a test image. Then, an image comparison was performed to mark similar blocks on the test image. The block that best resembled the template was chosen as the new template (or matching adaptive template). Afterwards, image subtraction was applied to subtract the pixels at the same location in each block of the test image from the matching adaptive template. The control limit law employed logic operation to separate the defect from the background region. The complete defect structure was obtained by the morphology method. Next, feature values, including defect gray value, red, green, and blue (RGB) color components, and aspect ratio were obtained as the classifier input. The experimental results showed that defect recognition could be completed as fast as 0.154s using the proposed recognition system and software. In micro-defect classification, back-propagation neural network (BPNN) and minimum distance classifier (MDC) served as the defect classification decision theories for the five acquired feature values. To validate the proposed system, this study used 41 defects as training samples, and treated the feature values of 307 test samples as the BPNN classifier inputs. The total recognition rate was 93.7%. When an MDC was used, the total recognition rate was 96.8%, indicating that the MDC method is feasible in applying automatic optical inspection technology to classify micro-defects of color filters. The proposed system is proven to successfully improve the production yield and lower costs.

Contamination Grades Recognition of Ceramic Insulators Using Fused Features of Infrared and Ultraviolet Images

In order to realize the non-contact measurement of ceramic insulator contamination severity, a method based on feature level fusion of infrared (IR) and ultraviolet (UV) image information is proposed in this paper. IR and UV images of artificially polluted insulators were obtained from high voltage experiments at 80%, 85% and 90% RH. After the preprocessing of images, IR and UV features were calculated, respectively. Then, feature selection based on Fisher criterion was adopted to gain features, which have the ability to distinguish different contamination grades effectively. In feature level fusion section, kernel principal component analysis (KPCA) was applied to the dimensionality reduction fusion of IR and UV features and obtain three-dimensional fused features. A particle swarm optimized back propagation neural network (PSO-BPNN) classifier was constructed and trained to recognize the contamination grades. Experimental results indicate that the feature level fusion of IR and UV information based on KPCA has capability to characterize the contamination grades comprehensively. Compared with recognition using IR or UV features separately, recognition based on the feature level fusion is more accurate and effective. This study provides a new methodology for the measurement of insulator contamination severity at working condition.

Automatic Characterization of Benign and Malignant Masses in Mammography

The paper aims to develop an automated breast mass characterization system for assisting the radiologist to analyze the digital mammograms. Mammographic Image Analysis Society (MIAS) database images are used in this study. Fuzzy C-means technique is used to segment the mass region from the input image. GLCM texture features namely contrast, correlation, energy and homogeneity are obtained from the region of interest. The texture features extracted from gray level co-occurrence matrix (GLCM) are computed at distance d=1 and θ=0o, 45o, 90o, 135o. These with three classifiers namely adaboost, back propagation neural network and sparse representation classifiers are used for characterizing the region containing either benign mass or malignant mass. The experimental results show the SRC classifier is more effective with an accuracy of 93.75% and with the Mathew's correlation coefficient (MCC) of 87.35%.

Grid Color Moment Features in Glaucoma Classification

Automated diagnosis of glaucoma disease is focused on the analysis of the retinal images to localize, perceive and evaluate the optic disc. Clinical decision support system (CDSS) is used for glaucoma classification in human eyes. This process depends mainly on the feature type that can be morphological or non-morphological. It is originated in the retinal image analysis technique that used color feature, texture features, extract structure, or contextual. This work proposes an empirical study on a narrative automated glaucoma diagnosis classification system based on both Grid Color Moment method as a feature vector to extract the color features (non-morphological) and neural network classifier. Consequently, these features are fed to the back propagation neural network (BPNN) classifier for automated diagnosis. The proposed system was tested using an open RIM-ONE database with accurate gold standards of the optic nerve head. This work classifies both normal and abnormal defected retina with glaucoma images. The experimental results achieved an accuracy of 87.47%. Thus, the proposed system can detect the early glaucoma stage with good accuracy.

A Machine Learning Framework for Detecting Landslides on Earthen Levees Using Spaceborne SAR Imagery

Earthen levees have a significant role in protecting large areas of inhabited and cultivated land in the United States from flooding. Failure of the levees can result in loss of life and property. Slough slides are among the problems which can lead to complete levee failure during a high water event. In this paper, we develop a method to detect such slides using X-band synthetic aperture radar (SAR) data. Our proposed methodology includes: radiometric normalization of the TerraSAR image using high-resolution digital elevation map (DEM) data; extraction of features including backscatter and texture features from the levee; a feature selection method based on minimum redundancy maximum relevance (mRMR); and training a support vector machine (SVM) classifier and testing on the area of interest. To validate the proposed methodology, ground-truth data are collected from slides and healthy areas of the levee. The study area is part of the levee system along the lower Mississippi River in the United States. The output classes are healthy and slide areas of the levee. The results show the average classification accuracies of approximately 0.92 and Cohen's kappa measures of 0.85 for both healthy and slide pixels using ten optimal features selected by mRMR with a sigmoid SVM. A comparison of the SVM performance to the maximum likelihood (ML) and back propagation neural network (BPNN) shows that the average accuracy of the SVM is superior to that of the BPNN and ML classifiers.

Tumor Disease Multiclass Prediction using Biomolecular Gene Expression Data by Signal Processing and Computational Intelligence Techniques

Tumor disease multiclass prediction from nucleotide expression is an emerging research area in the field of bioinformatics. Gene expression profiling has been emerged as an efficient technique for cancer classification as well as for diagnosis, prognosis, and treatment purposes. Studying cancer microarray gene expression data is a challenging task because microarray is high dimensional dataset with a noisy data. Efficient feature extraction and computational method development is indispensible for the analysis. In this paper a feature extraction method by Discrete Cosine Transform (DCT) and discrete wavelet transform (DWT) has been proposed to detect informative genes effectively. DCT offers a dimension reduction in feature sets. Again the approximation coefficients obtained by the decomposition at a level in DWT is used as the features for further study. Then K-means algorithm is applied on optimized feature datasets to cluster. These cluster information are feature sets and classified using Back-Propagation Neural Network (BPNN) classifier to efficiently predict the class. The potential of the proposed approach is validated by many benchmark datasets such as lungs cancer dataset, breast cancer data set, Prostate cancer dataset, brain tumor dataset, colon dataset, Leukemia dataset. The experimental results show that the proposed method can be a useful approach for cancer classification with low computational complexity and high accuracy.

Automatic Classification of Remote Sensing Images Using Multiple Classifier Systems

It is a challenge to obtain accurate result in remote sensing images classification, which is affected by many factors. In this paper, aiming at correctly identifying land use types reflec ted in remote sensing images, support vector machine, maximum likelihood classifier, backpropagation neural network, fuzzy c-means, and minimum distance classifier were combined to construct three multiple classifier systems (MCSs). Two MCSs were implemented, namely, comparative major voting (CMV) and Bayesian average (BA). One method called WA-AHP was proposed, which introduced analytic hierarchy process into MCS. Classification results of base classifiers and MCSs were compared with the ground truth map. Accuracy indicators were computed and receiver operating characteristic curves were illustrated, so as to evaluate the performance of MCSs. Experimental results show that employing MCSs can increase classification accuracy significantly, compared with base classifiers. From the accuracy evaluation result and visual check, the best MCS is WA-AHP with overall accuracy of 94.2%, which overmatches BA and rivals CMV in this paper. The producer’s accuracy of each land use type proves the good performance of WA-AHP. Therefore, we can draw the conclusion that MCS is superior to base classifiers in remote sensing image classification, and WA-AHP is an efficient MCS.

Wind Power Forecasting in a Residential Location as Part of the Energy Box Management Decision Tool

A number of location characteristics, such as buildings, mountains, and trees, are likely to influence the wind flow that reaches a microwind turbine located at a residential area, and as a consequence they may affect the actual wind speed that is potentially utilized by the turbine. In this context, simple regional predictions for the wind energy from the nearest location available can easily lead to unacceptable modeling errors. There is thus a need to develop a framework for predicting the values of the wind speed at the desired location. This work addresses the aforementioned issue by employing a multilayer feed-forward back-propagation neural network for classification that utilizes the global forecast system (GFS) predictions on wind speed and direction to identify patterns of the wind behavior at the location considered in order to obtain a stochastic distribution of the daily wind speed. The proposed approach aims to support the implementation of an enhanced energy box (EB) management decision tool, while its feasibility is demonstrated through a case example for a region in the south of Portugal.

An automatic inspection system for segment-type display module defects

Purpose – This study aimed to developed a defect detection system for a segment-type display module panel. Design/methodology/approach – The system included a data acquisition card, a video camera, a computer and a display module on a testing table. The video camera captured the display pattern of the display module and transferred it to the computer through the data acquisition card. The dynamic multi-thresholding method and analysis as well as back propagation neural network classification was used to classify the detected defects. Findings – The threshold values for the brightness at different positions in the display module image were obtained using the neural network and then stored in the look-up table, using two to six matrixes. Originality/value – The recognition speed was faster and the system was more flexible in comparison to the previous system. The proposed method, using unsophisticated and economical equipment, was also verified as providing highly accurate results with a low error rate.

Use of Hyperspectral Imaging to Discriminate the Variety and Quality of Rice

This paper investigated the use of hyperspectral imaging (HSI) to discriminate the variety and quality of rice. Hyperspectral images (400–1,000 nm) of paddy rice samples were acquired to extract both spectral and image information. Dimension reduction was carried out on the region of interest (ROI) of the images by principal component analysis (PCA). The first principal components (PCs) explained over 98 % of variances of all spectral bands. Chalkiness degree and shape feature (‘MajorAxisLength’, ‘MinorAxisLength’, length-width ratio, ‘Perimeter’ and ‘Eccentricity’) were further extracted and used for subsequent rice variety discrimination by PCA and back propagation neural network (BPNN). An integration of spectral and image data was used for BPNN classification. The BPNN model based on spectral data (seven optimal wavelengths) achieved better results than PCA based on spectral data (seven optimal wavelengths) in variety discrimination with classification accuracy of 89.18 and 89.91 % for PCA and BPNN model, respectively. The BPNN model based on data fusion achieved the best results (94.45 %), which was superior to the results based on spectral data (89.91 %) or image data (88.09 %) alone. Finally, the resulting classification maps were able to visualize different rice varieties. The results demonstrated that discrimination of rice variety and quality with HSI technology was feasible and could be utilized for quality control purposes and/or for innovative sorting of rice.

Improving activated sludge classification based on imbalanced data

A fast and accurate classification method for sewage sludge biological activity classification is of great significance for wastewater treatment. However, the data are often imbalanced and the accuracy of traditional classification algorithms applied to imbalanced small classes of data is very low. Such small classes are crucial application data. Therefore, based on the analysis of eight microorganisms, a novel method is proposed in this paper for the classification of activated sludge known as balanced support-vector-based back-propagation (SV-BP) neural network. It first splits the multiclass classification problem into a plurality of pairwise classification problems and uses a support vector machine (SVM) to achieve equalization. Second, the new dataset is produced, following which back-propagation neural network (BPNN) is used for training and classification. To examine the efficiency of the model, 1731 real data points are collected from a wastewater treatment factory and divide the data into four classes with the help of wastewater experts. Based on the new model, data redundancy and noise are greatly reduced. With area under the curve (AUC) measurements, we find that the AUC of SV-BP is 6.9% higher than classical BPNN. In addition, the small-class recognition rate of SV-BP is far better than that by classical BPNN and SVM algorithms.

Visual Character N-Grams for Classification and Retrieval of Radiological Images

Diagnostic radiology struggles to maintain high interpretation accuracy. Retrieval of past similar cases would help the inexperienced radiologist in the interpretation process. Character n-gram model has been effective in text retrieval context in languages such as Chinese where there are no clear word boundaries. We propose the use of visual character n-gram model for representation of image for classification and retrieval purposes. Regions of interests in mammographic images are represented with the character ngram features. These features are then used as input to back-propagation neural network for classification of regions into normal and abnormal categories. Experiments on miniMIAS database show that character n-gram features are useful in classifying the regions into normal and abnormal categories. Promising classification accuracies are observed (83.33%) for fatty background tissue warranting further investigation. We argue that Classifying regions of interests would reduce the number of comparisons necessary for finding similar images from the database and hence would reduce the time required for retrieval of past similar cases.