Back Propagation Neural Network Classifier Research Articles

Kannada Handwritten numeral Recognition using FFBPNN Classifiers

This paper presents the performance of Kannada handwritten numeral recognition using feed forward back propagation neural network (FFBPNN) classifiers. The classifier is designed to recognize the Kannada handwritten numerals. Samples are represented by the few features extracted by the zoning technique. The input numeral samples in binary form are stored in a fixed window size of 12x12 and partitioned into nine sub regions of 4x4 sizes for their representation. A normalized feature value is computed by the one‟s present each sub region for their representation in two different approaches. On experimentation, it is found the overall recognition rate of 99.7% and 95.5% for the feature extraction approaches M1 and M2 respectively. General Terms Pattern Recognition, Feed Forward Back Propagation Neural Network, Character Recognition

Research of Defect Detection Method Noise for Bevel Gear

Noise signal analysis method is widely available for gearbox bevel gear fault detection. However, the noise from the gearbox is usually concealed by background noise, which leads to poor efficiency analysis. This paper reports an ensemble empirical mode decomposition (EEMD) and neural network method for bevel gear fault detection. To extract useful signal, EEMD algorithm was firstly applied to get rid of the background noise. Characteristics from a group of discriminating defect status were then chosen to build the eigenvector. Finally, the eigenvector was imported into a back propagation (BP) neural network classifier for defect diagnosis automatically. Experimental results show that the proposed approach is capable for signal denoising and providing distinguishing characteristics of founded fault. The developed method is an accurate approach to detect fault for tested bevel gear.

Analysis of GLCM Feature Extraction for Choosing Appropriate Angle Relative to BP Classifier

GIS can manage remotely sensed images, users must have an appropriate digital map that represents lands each one has information according to its owner, status, and some other data. The classification of such lands is a great problem which take long time depending on human efforts. Many kinds of classifications had been used , one of them is the use of supervised multi-layer perceptron with backpropagation neural network classifier and using second order statistics Gray Level Co-occurrence Matrix (GLCM) to calculate eight textural features for each one of three visible bands (RGB) for each land sample. In this research we analyzed the GLCM feature extraction algorithm to detect the appropriate angle that can be chosen , relatively with the training of BP classifier had been used according to the number of hidden nodes inside the hidden layer of ANN . As a result the system produce high accuracy with the best angle choosing of GLCM , these results are achieved by comparing the classification results from system test trials with desired user predefined

R-Peak Detection using Daubechies Wavelet and ECG Signal Classification using Radial Basis Function Neural Network

This paper employed the Daubechies wavelet transform (WT) for R-peak detection and radial basis function neural network (RBFNN) to classify the electrocardiogram (ECG) signals. Five types of ECG beats: normal beat, paced beat, left bundle branch block (LBBB) beat, right bundle branch block (RBBB) beat and premature ventricular contraction (PVC) were classified. 500 QRS complexes were arbitrarily extracted from 26 records in Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) arrhythmia database, which are available on Physionet website. Each and every QRS complex was represented by 21 points from p1 to p21 and these QRS complexes of each record were categorized according to types of beats. The system performance was computed using four types of parameter evaluation metrics: sensitivity, positive predictivity, specificity and classification error rate. The experimental result shows that the average values of sensitivity, positive predictivity, specificity and classification error rate are 99.8%, 99.60%, 99.90% and 0.12%, respectively with RBFNN classifier. The overall accuracy achieved for back propagation neural network (BPNN), multilayered perceptron (MLP), support vector machine (SVM) and RBFNN classifiers are 97.2%, 98.8%, 99% and 99.6%, respectively. The accuracy levels and processing time of RBFNN is higher than or comparable with BPNN, MLP and SVM classifiers.

An Intelligent System Based on Back Propagation Neural Network and Particle Swarm Optimization for Detection of Prostate Cancer from Benign Hyperplasia of Prostate

Conventional clinical diagnostic methods are generally based on a single classifier. In present paper, we propose a hybrid Backpropagation neural network (BPNN) classifier based particle swarm optimization (PSO) method. In the present paper by combining the principles two algorithm, we propose a new but simple hybrid algorithm called BPNN_ PSO. Our novel algorithm optimizes BPNN with PSO and reduces computational time of the training phase of BPNN. The performance of the algorithm has been tested with prostate cancer. A total of 360 medical records collected from the patients suffering from neoplasia diseases have been used to train and test the proposed algorithm. The results show that the proposed BPNN–PSO algorithm can achieve very high diagnosis accuracy (98%) and it proving its usefulness in supporting of clinical decision process of prostate cancer. Comparing the simulated results of the above two cases, training the neural network by PSO technique gives more accurate (in terms of sum square error) and also faster (in terms of number of iterations and simulation time) results than BPNN. By using these hybrid method for building machine learning classifiers, we can significantly improve diagnostic performance with respect to the results of clinical practice

A Robust Environmental Sound Recognition System using Frequency Domain Features

In ubiquitous environments, analysis and classification of sound plays a critical role in various acoustic-based recognition systems. This work aims to contribute towards building an automatic sound recognition system that can understand the surrounding environment by the audio information. In this paper, an acoustic signal based context awareness system is proposed for detecting sound events in five different real-world environment.This approach is based on Back Propagation Neural Network (BPNN) classifier using a new feature set from frequency-domain features. The experiments on various categories illustrate that the results of recognition are significant and effective.

A New Back-Propagation Neural Network Algorithm for a Big Data Environment Based on Punishing Characterized Active Learning Strategy

This paper introduces the active learning strategy to the classical back-propagation neural network algorithm and proposes punishing-characterized active learning Back-Propagation (BP) Algorithm (PCAL-BP) to adapt to big data conditions. The PCAL-BP algorithm selects samples and punishments based on the absolute value of the prediction error to improve the efficiency of learning complex data. This approach involves reducing learning time and provides high precision. Numerical analysis shows that the PCAL-BP algorithm is superior to the classical BP neural network algorithm in both learning efficiency and precision. This advantage is more prominent in the case of extensive sample data. In addition, the PCAL-BP algorithm is compared with 16 types of classical classification algorithms. It performs better than 14 types of algorithms in the classification experiment used here. The experimental results also indicate that the prediction accuracy of the PCAL-BP algorithm can continue to increase with an increase in sample size.

Weaver Ant Colony Optimization-Based Neural Network Learning for Mammogram Classification

Neural Networks (NNs) have been efficaciously used for classification purposes in medical domains, including the classification of microcalcifications in digital mammograms. Unfortunately, for a NN to be effective in a particular purview, its architecture, training algorithm and the domain variables selected as inputs must be amply chosen. In this paper, a novel Ant Colony Optimization (ACO) based learning approach with a modified architecture is proposed to speed up the learning phase of a Backpropagation Neural Network (BPN) classifier. The novel ACO simulates the behavior of weaver ants, known for their unique nest building behavior where workers construct nests by weaving together leaves using larval silk. The proposed Weaver Ant Colony Optimization (WACO) based Backpropagation Neural Network (WACO-BPN) is applied for classifying digital mammograms received from MIAS database. The performance is analyzed with Receiver Operating Characteristics (ROC) curve. The greater accuracy of 97% states the grander performance of the proposed neural network learning approach.

Classification of Normal and Myopathy EMG Signals using BP Neural Network

Electromyography (EMG) signal is the muscle electrical activity. Electromyography is a technique for detecting and recording the electrical potential generated by muscle cells. This EMG signals are used in medical professionals to determine specific disorders. This paper basically deals with the analysis of different electromyography signals (NOR & MYO). In this paper, new method for classification of myopathy patient’s and healthy subjects with the help of EMG signal by using back propagation neural network classifier are proposed. This methodology provided 96.75 % accuracy in classification of Myopathy and normal EMG signals.

A comparative study of computer-aided classification systems for focal hepatic lesions from B-mode ultrasound

A comparative study of three computer-aided classification (CAC) systems for characterization of focal hepatic lesions (FHLs), such as cyst, hemangioma (HEM), hepatocellular carcinoma (HCC) and metastatic carcinoma (MET), along with normal (NOR) liver tissue is carried out in the present work. In order to develop efficient CAC systems a comprehensive and representative dataset consisting of B-mode ultrasound images with (1) typical and atypical cases of cyst, HEM and MET lesions, (2) small and large HCC lesions and (3) NOR liver cases have been used for designing K-nearest neighbour (KNN), probabilistic neural network (PNN) and a back propagation neural network (BPNN) classifiers. For differential diagnosis between atypical FHLs, expert radiologists often visualize the textural characteristics of regions inside and outside the lesion. Accordingly in the present work, texture features and texture ratio features are computed from regions inside and outside the lesions. A feature set consisting of 208 texture features (i.e. 104 texture features and 104 texture ratio features) is subjected to principal component analysis (PCA) for dimensionality reduction; it is observed that maximum accuracy of 87.7% is obtained for a PCA-BPNN-based CAC system in comparison to 86.1% and 85% as obtained by PCA-PNN and PCA-KNN-based CAC systems. The sensitivity of the proposed PCA-BPNN based CAC system for NOR, Cyst, HEM, HCC and MET cases is 82.5%, 96%, 93.3%, 90% and 82.2%, respectively. The sensitivity values with respect to typical, atypical, small HCC and large HCC cases are 85.9%, 88.1%, 100% and 87%, respectively. Keeping in view the comprehensive and representative dataset used for designing the classifier, the results obtained by the proposed PCA-BPNN-based CAC system are quite encouraging and indicate its usefulness to assist experienced radiologists for interpretation and diagnosis of FHLs.

Automatic defect inspection system of colour filters using Taguchi-based neural network

A coloured filter is a critical part of an LCD panel, especially to present a high quality colour display. At present, the defect detection of colour filters is conducted by manual inspection in the final product stage. However, poor detection efficiency and subjective judgment of manual inspection undermine accuracy. Therefore, this study applied image processing technology and the neural network to detect surface defects of colour filters in order to prevent losses arising from incorrect detection, lower production costs, and effectively improve yield. A back-propagation neural network (BPNN) classifier was selected to train the features. The results showed that the proposed method can be successfully applied in defect detection of colour filters to reduce artificial detection errors. In addition, the Taguchi method was used with BPNN to save time searching optimal learning parameters by the trial and error method, which achieves faster convergence, smaller convergent errors and better recognition rate. The results proved that the root-mean-square error (RMSE) of the Taguchi-based BPNN at final convergence is 0.000254, and recognition rate reaches 94%. Therefore, the proposed method has good effects in detecting the micro defects of a colour filter panel.

An Empirical Approach to Optimize Design of Backpropagation Neural Network Classifier for Textile Defect Inspection

Automated fabric inspection systems have been drawing plenty of attention of the researchers in order to replace manual inspection. Two difficult problems are mainl y posed by automated fabric inspection systems. They are defect detection and defect classification. Backpropagation is a popular learning algorithm and very promising for defect classification. In general, works reported to date have claimed varying level of successes in detection and classification of different types of defects through backpropagation model. In those published works, no investigation has been reported regarding to the variation of major performance parameters of neural network (NN) based classifiers such as training time and classification accuracy based on network topology and training parameters. As a result, application engineer has little or no guidance to take design decisions for reaching to optimum structure of NN based defect classifiers in general and backpropagation based in particular. Our work focuses on empirical investigation of interrelationship between design parameters and performance of backpropagation based classifier for textile defect classification. It is believed that such work will be laying the ground to empower application engineers to decide about optimum values of design parameters for realizing most appropriate backpropagation based classifier.

Multimodal Biometric System Using Finger Knuckle and Nail: A Neural Network Approach

Dorsum of the hand can be very useful in Personal identification but yet it has not that much extensive attention. Single scan of dorsum hand can give two biometric traits finger-knuckle and finger nail. This paper presents an approach to combine Finger-knuckle and finger-nail features. Finger nail biometric is considered as quite unique biometric trait hence we combine this trait with finger knuckle. We also developed an algorithm to extract ROI from Finger-knuckle and finger-nail. Finger- knuckle features are extracted using Mel Frequency Cepstral Coefficient (MFCC) technique and the features of finger-nail are extracted from second level wavelet decomposition. We combined these features using feature level fusion and feed forward back-propagation Neural Network for classification. The performance of the system has been tested on our own KVKR- knuckle database that includes 100 subject's dorsal hands. Evaluation results shows that increase in training set gives increased performance rate. The best performance of the proposed system reaches up to 97% with respective training of 90% of total dataset. 

Classification of benign and malignant brain tumor CT images using wavelet texture parameters and neural network classifier

Computational methods are useful for medical diagnosis because they provide additional information that cannot be obtained by simple visual interpretation. As a result, an enormous amount of computer vision research effort has been targeted at achieving automated medical image analysis. In this paper, we present the combination of wavelet statistical texture features (WST) obtained from two-level discrete wavelet-transformed (DWT) images and wavelet co-occurrence texture features (WCT) obtained from two-level DWT detail images for the classification of abnormal brain tissues into benign, malignant tumor of CT images. Our proposed system consists of four phases: (1) segmentation of region of interest, (2) discrete wavelet decomposition, (3) feature extraction and feature selection, and (4) classification and evaluation. The support vector machine is employed to segment the shape of tumor information. A combination of both WST and WCT texture features is extracted from tumor region of two-level discrete wavelet-transformed images. Genetic algorithm (GA) is used to select the optimal texture features from the set of extracted features. The probabilistic neural network classifier (PNN) is built to classify the abnormal brain tissues into benign, malignant tumor images and evaluate the performance of classifier by comparing the classification results of the PNN classifier with linear vector quantization (LVQ) neural network classifier, back propagation neural network (BPN) classifier. The results of PNN, LVQ, BPN classifiers for the texture analysis methods are evaluated using statistical analysis and receiver operating characteristic analysis. From the experimental results, it is inferred that the best classification performance is achieved by PNN than LVQ and BPN classifiers. The system has been tested with real data of 80 benign, malignant CT brain tumor images and has achieved satisfactory results.

Detection for Corn/Weed Images Using Moment Invariants by BPNN Classifier

This study was conducted to discriminate the weed from the corn in a field combined neural network classifier with image processing technology. The corn and weed images were scanned using a colour imaging system. In the first step, an approximate location of the object of interest was determined by minimum enclosing rectangle, in which image processing was done to obtain the binary image. In the second step, the seven invariant moments were extracted from binary images and used as input to the back propagation neural network (BPNN) classifier. The training set was used to construct shape model representing the objects. The detection accuracy was enhanced by adjusting the number of neurons in the network. Experimental results showed that the BPNN classifier achieved overall detection accuracy of 94.52% with 7-28-1.

Species classification of aquatic plants using GRNN and BPNN

Computer-aided plant species identification acts significantly on plant digital museum system and systematic botany, which is the groundwork for research and development of plants. This work presents a method for plant species identification using the images of flowers. It focuses on the stable feature extraction of flowers such as color, texture and shape features. Color-based segmentation using k-means clustering is used to extract the color features. Texture segmentation using texture filter is used to segment the image and obtain texture features. Sobel, Prewitt and Robert operators are used to extract the boundary of image and to obtain the shape features. From 405 images of flowers, color, texture and shape features are extracted. Classification of the plants into dry land plants and aquatic plants, the aquatic plant species into wet and marsh aquatic plants, wet aquatic plants into Iridaceae and Epilobium family and marsh aquatic plants into Malvaceae and Onagraceae family, the Iridaceae family is again classified into Babiana and Crocus species, the family Epilobium into Canum and Hirsutum, the family Malvaceae into Mallow and Pavonia, the family Onagraceae into Fuschia and Ludwigia species are done using general regression neural network and backpropagation neural network classifiers.

Performance Evaluation of Classifiers for Detection of Alcoholics Using Electroencephalograms (EEG)

This paper suggests an automated detection of alcoholics and controls using Electroencephalograms (EEG). EEG signals are recorded from healthy and alcoholic subjects during visual object recognition task. Threshold of 100 �V is used to remove the eye blink artifact and the gamma sub band (30–50 Hz) is extracted using elliptic band pass filter of 6th order. Different classical (spectral entropy), model based (power spectral density (PSD) estimation by Burg autoregressive (AR) model and eigenvector PSD estimation methods (MUSIC) in frequency domain are used as feature selection parameters. Three classification models such as feed forward Back propagation neural network (BPNN), Probabilistic neural network (PNN) and Support Vector Machine (SVM) networks are used to classify alcoholics and control subjects. The performances of BPNN and PNN classifiers are evaluated using 10-fold cross validation and holdout cross validation is performed to evaluate SVM classifier. Results show that model based and eigenvector based feature extraction methods are more accurate and sensitive than the conventional method such as spectral entropy estimation method to discriminate alcoholics from healthy controls. The SVM and PNN classifiers perform better than the BPNN classifier in terms of classification accuracy and computational complexity. However to achieve high classification accuracy, SVM does not require either the preprocessing of the input signal as in BPNN classifier or the tuning of the spread factor as in PNN classifier. Hence among the three classifiers used in this work, SVM seems to be an optimal classifier for classifying alcoholics and healthy subjects.

ON LAND SLIDE DETECTION USING TERRASAR-X OVER EARTHEN LEVEES

Abstract. Earthen levees have an important role to protect large areas of inhabited and cultivated land in the US from flooding. Failure of the levees can threaten the loss of life and property. One of the problems which can lead to a complete failure during a high water event is a slough slide. In this research, we are trying to detect such slides using X-band SAR data. Our methodology consists of the following four steps: 1) segmentation of the levee area from background; 2) extracting features including backscatter features and texture features; 3) training a back propagation neural network classifier using ground-truth data; and 4) testing the area of interest and validation of the results using ground truth data. A dual-polarimetric X-band image is acquired from the German TerraSAR-X satellite. Ground-truth data include the slides and healthy area. The study area is an approximately 1 km stretch of levee along the lower Mississippi River in the United States. The output classification shows the two classes of healthy and slide areas. The results show classification accuracies of approximately 67% for detecting the slide pixels.

Facial Expression Recognition System using Statistical Feature and Neural Network

In this paper, a new technique for facial expression recognition is proposed which uses the statistical feature of the whole face and classify the expression using neural network classifier. When the face image is input, region of interest (ROI) is being obtained to evaluate the statistical feature of the face. Using these, features we classify the face into one of the seven different expressions by using multi label Back Propagation neural network classifier. To demonstrate the proposed recognition technique we use JAFFE facial database and the whole program is being implemented in MATLAB 7.0. General Terms Pattern Recognition.

Landsat-TM-Based Discrimination of Lithological Units Associated with the Purtuniq Ophiolite, Quebec, Canada

In order to better constrain the utility of multispectral datasets in the characterization of surface materials, Landsat Thematic Mapper (TM) data were evaluated in the discrimination of geological classes in the Cape Smith Belt of Quebec, a greenstone belt that hosts Early Proterozoic units including those of the Purtuniq ophiolite. Ground-based measurements collected for the study area highlight the importance of chemical alteration in controlling the reflectance properties of key geological classes. The spatial distribution of exposed lithologies in the study area was determined through (1) image classification using a feedforward backpropagation neural network classifier; and (2) generation of fraction images for spectral end members using a linear unmixing algorithm and ground reflectance data. Despite some shortcomings, the database of surface cover generated by the neural network classifier is a useful representation of the spatial distribution of exposed geological materials in the study area, with an overall agreement with ground truth of 87.7%. In contrast, the fraction images generated through unmixing are poor representations of ground truth for several key lithological classes. These results underscore both the considerable utility and marked limitations of Landsat TM data in the mapping of igneous and metamorphic lithologies.