Back Propagation Neural Network Classifier Research Articles

Ontology based multiobject segmentation and classification in sports videos

The primary objective is to identify and segments the multiple, partly occluded objects in the image. The subsequent stage carry out our approach, primarily start with frame conversion. Next in the preprocessing stage, the Gaussian filter is employed for image smoothening. Then from the preprocessed image, Multi objects are segmented through modified ontology-based segmentation, and the edge is detected from the segmented images. After that, from the edge detected frames area is extracted, which results in object detected frames. In the feature extraction stage, attributes such as area, contrast, correlation, energy, homogeneity, color, perimeter, circularity are extorted from the detected objects. The objects are categorized as human or other objects (bat/ball) through the feed-forward back propagation neural network classifier (FFBNN) based upon the extracted attributes.

Feature Extraction of Surface Electromyography Using Wavelet Weighted Permutation Entropy for Hand Movement Recognition.

The feature extraction of surface electromyography (sEMG) signals has been an important aspect of myoelectric prosthesis control. To improve the practicability of myoelectric prosthetic hands, we proposed a feature extraction method for sEMG signals that uses wavelet weighted permutation entropy (WWPE). First, wavelet transform was used to decompose and preprocess sEMG signals collected from the relevant muscles of the upper limbs to obtain the wavelet sub-bands in each frequency segment. Then, the weighted permutation entropies (WPEs) of the wavelet sub-bands were extracted to construct WWPE feature set. Lastly, the WWPE feature set was used as input to a support vector machine (SVM) classifier and a backpropagation neural network (BPNN) classifier to recognize seven hand movements. Experimental results show that the proposed method exhibits remarkable recognition accuracy that is superior to those of single sub-band feature set and commonly used time-domain feature set. The maximum recognition accuracy rate is 100% for hand movements, and the average recognition accuracy rates of SVM and BPNN are 100% and 98%, respectively.

Comparative analysis of texture feature extraction techniques for rice grain classification

Classifications of eight different varieties of rice grain are discussed in this study based on various texture models. Four local texture feature extraction techniques are proposed and three sets of texture features (SET-A, SET-B and SET-C) are formed, for the classification task. Performances of the proposed feature sets are compared with the existing techniques based on, run length matrix, co-occurrence matrix, size zone matrix, neighbourhood grey tone difference matrix and wavelet decomposition, towards classification of rice grain using a back propagation neural network (BPNN). The proposed techniques are also tested against publicly available data from Brodatz's texture data set and their results are compared with other techniques. The classification accuracy by the BPNN classifier is also compared with other statistical classifiers namely, K-nearest neighbour, linear discriminant classifier and Naive Bayes classifier. It is found that, the proposed feature sets yield better classification results on both rice data and Brodatz's data. Results show that, feature SET-B, is able to classify rice grain with an average classification accuracy of 99.63% with a minimum of six features.

Method to determine nuclear accident release category via environmental monitoring data based on a neural network

After a severe nuclear accident, the source term is typically unknown. Therefore, great importance is attached to obtaining source term information for subsequent emergency planning. The purpose of the study involved performing a rapid determination of the release category using the gamma dose rate monitoring data over a short period during an accident. The release categories PWR1–PWR9 in 'The Reactor Safety Study' of the United States were used as a reference. The International Radiological Assessment System (InterRAS) was used to construct a nuclear accident model and generate the required simulation data. After a series of experiments, appropriate parameters were selected to construct the backpropagation neural network (BPNN) classifier to estimate the release category. The genetic algorithm (GA) and simulated annealing (SA) algorithm were used to search for the weights and thresholds of the BPNN classifier in advance, and this avoided the problem wherein bad initial values can cause the classifier to fall into local minimums, decreased training time, and improved prediction accuracy from 98.36% to 99.10%. With respect to the possible absence of gamma dose rate monitoring data, particle swarm optimisation (PSO) was used to complete the missing data, thereby ensuring that the classifier can normally predict the release category. After testing, in the absence of 4 of the total 16 gamma dose rate data, the classifier can still maintain a prediction accuracy of more than 80%.

ARTIFICIAL INTELLIGENT SYSTEM FOR SKIN DISEASES CLASSIFICATION

In this paper, an artificial intelligent technique is proposed for skin disease detection and classification. The suggested method comprises four stages, including segmentation, extraction of textural features, and classification. The stretch-based enhanced algorithm has been adapted for image enhancement. Then the method of an active contour is used for segmentation to determine the skin lesion in tissue. Textural features are obtained from the segmented skin lesion. As several numbers of the features can affect the classification precision, ideal feature selection is made to exclude features that are less informative and unnecessary. The feature selection is adjusted with a regularized random forest. Finally, the classification algorithms by support vector machine and a back-propagation neural network (BPNN) are implemented. The dataset consists of 400 dermoscopic images in total divided into 200 benign and 200 malignant skin diseases extracted from the dermoscopic images PH2 database. The result of detecting and classifying the dermoscopic images on these images yielded an accuracy of 99.7%, a sensitivity of 99.4%, and a specificity of 100% by BPNN. The experiential results confirmed that the BPNN classifier is best rather than an SVM classifier for skin disease images. This proposed model will be advanced to support the skin image processing techniques that provided a more accurate diagnosis and rapid treatment plan.

Effect of Data Normalization on Accuracy and Error of Fault Classification for an Electrical Distribution System

ABSTRACT Unpredictable occurrence of electrical faults on a distribution system leads to power outages for electricity consumers. This makes accurate detection and classification of faults an important task. To isolate faulty sections, output of artificial intelligence (AI)based fault classifier is used to design optimum protection schemes for distribution systems. To enhance classifier’s performance, digital signal processing techniques like discrete wavelet transform (DWT) are implemented to extract features from fault signals. If extracted features have large variations, normalization is required to convert them on a common scale. Here, the error and accuracy in classification of line to line fault current (L-L fault) is evaluated if z-score normalized and not normalized inputs are used for backpropagation neural network (BPNN) classifier. Experiments are conducted on IEEE 13 bus test distribution system simulated in MATLAB. Results clearly show significant increase in accuracy of fault classification with considerable reduction in error to classify the faults due to normalization of input data.

Isolated word language identification system with hybrid features from a deep belief network

SummaryThe representation of good audio features is the first and foremost requirement for improving the identification performance of any system. Most of the representation learning approaches are based on connectionist systems to learn and extract latent features from the speech data. This research work presents a hybrid feature extraction approach to integrate Mel‐Frequency Cepstral Coefficients (MFCC) features with Shifted Delta Cepstral (SDC) coefficients features, which are further stacked to Deep Belief Network (DBN), for yielding new feature representations of the speech signals. DBN is utilized for unsupervised feature learning on the extracted MFCC‐SDC acoustic features. A 3‐layer Back Propagation Neural Network (BPNN) classifier is initialized in terms of the learning outcomes of hidden layers of DBN for identifying language from the uttered speech. The efficiency of the proposed approach is evaluated by simulating several experimental algorithms on the user‐defined database of isolated words in four languages, namely, Tamil, Malayalam, Hindi, and English, in the working platform of MATLAB. The obtained results for the proposed hybrid approach MFCC‐SDC‐DBN are promising. The proposed approach is also compared with the baseline feature extraction approach MFCC‐SDC by utilizing traditional acoustic features and BPNN classifier. The accuracy obtained with our proposed approach is 98.1% whereas that of the baseline approach is 82%, thereby providing an overall improvement of 16.1%.

Analysis of journal evaluation indicators: an experimental study based on unsupervised Laplacian score

Academic journal rankings have always been a hot topic because the journals ranking will be significantly affected by using different ranking systems or bibliometrics indicators. Many studies have shown that feature selection methods are beneficial to categorize and rank journals according to the different evaluation indicators. However, such methods are limited to analyzing the journals published in a single discipline. In this article, we conduct an experiment based on unsupervised Laplacian score to analyze journal evaluation indicators between three different disciplines (Biology, Artificial Intelligence, and Mathematics). The journal’s features considered in the study include: JIF, 5-Year JIF, CiteScore, SJR, SNIP, and H-index. Based on the results of the feature selection, we first use the spectral clustering technique to verify the impact of different feature combinations on journals categorization. Furthermore, k-Nearest Neighbor (k-NN) and Back Propagation Neural Network classifiers are trained to measure the classification performance by using each single feature or feature subsets. The experimental results demonstrate that the two classifiers achieve the superior classification performance when the input of the classifier is the set of JIF, CiteScore, and H-index. Finally, we design a subjective questionnaire to collect the opinion of the volunteers with regard to the six evaluation indiactors in terms of familiarity, complexity, and usability. According to the vote results of a seven-point Likert scale, JIF and H-index receive the relatively higher ratings in comparison to other bibliometric indicators.

An experimental study of stunned state detection for broiler chickens using an improved convolution neural network algorithm

Effective recognition method of broiler stunned state has always been an important issue in real industries. In recent years, recognition methods such as neural networks have been receiving increasing attention due to their great merits of high diagnostic accuracy and easy implementation. To improve the accuracy and efficiency of broiler stunned state recognition, an improved fast region-based convolutional neural network (You Only Look Once + Multilayer Residual Module (YOLO + MRM)) algorithm was proposed and applied to the recognition of three broiler stunned states: insufficient, appropriate and excessive stuns. The images were collected from a broiler-slaughtering line using a complementary metal-oxide semiconductor (CMOS) camera. The area of the head and wings of a broiler in the original image was marked according to the PASCAL VOC data format and the dataset of each broiler stunned state was obtained. The results showed that the YOLO + MRM algorithm achieved good performance with an accuracy of 96.77%. To compare YOLO + MRM with other models, similar experiments were conducted using a conventional back propagation neural network (BP-NN) classifier, as well as YOLO, and the recognition accuracies were 90.11% and 94.74%, respectively. YOLO + MRM can complete the detection task of more than 180,000 broilers per hour. Compared with the traditional method, little prior expertise on image recognition is required, the recognition accuracy and speed are improved obviously. This study has provided a foundation and highlighted the potential for automatically detecting the stunned state of broiler chickens, which is crucial for the success of an automatic electric stunning process in the poultry industry.

A Hybrid Genetic-Neuro Algorithm for Cloud Intrusion Detection System

The security for cloud network systems is essential and significant to secure the data source from intruders and attacks. Implementing an intrusion detection system (IDS) for securing from those intruders and attacks is the best option. Many IDS models are presently based on different techniques and algorithms like machine learning and deep learning. In this research, IDS for the cloud computing environment is proposed. Here in this model, the genetic algorithm (GA) and back propagation neural network (BPNN) is used for attack detection and classification. The Canadian Institute for Cyber-security CIC-IDS 2017 dataset is used for the evaluation of performance analysis. Initially, from the dataset, the data are preprocessed, and by using the genetic algorithm, the attack was detected. The detected attacks are classified using the BPNN classifier for identifying the types of attacks. The performance analysis was executed, and the results are obtained and compared with the existing machine learning-based classifiers like FC-ANN, NB-RF, KDBN, and FCM-SVM techniques. The proposed GA-BPNN model outperforms all these classifying techniques in every performance metric, like accuracy, precision, recall, and detection rate. Overall, from the performance analysis, the best classification accuracy is achieved for Web attack detection with 97.90%, and the best detection rate is achieved for Brute force attack detection with 97.89%.

A Digital Signage Audience Classification Model Based on the Huff Model and Backpropagation Neural Network

Digital signage is an important outdoor advertising medium in cities. However, advertising on digital signage often lacks pertinence. Thus, it is important to introduce an accurate digital signage audience classification method to facilitate targeted advertising. In this study, a multi-label classification model based on a backpropagation (BP) neural network and the Huff model, referred to as the Huff-BP model, is proposed to investigate digital signage audience classification. A case study is performed on outdoor digital signage within the 6th Ring Road in Beijing, China, and economic census, population census, average housing price, social media check-in and the centrality of traffic networks as research data. The data are divided into 100 × 100-1,000 × 1,000 m normal grids. Multi-label classification modelling factors for various grid scales are constructed. The BP neural network classification algorithm is improved to solve the multilabel classification problem. In addition, an improved Huff model is used to calculate the digital signage influence values between each grid cell and integrated into the improved BP neural network to classify modelling factors at various scales. Finally, four metrics are used to examine the effectiveness of the proposed model. The results show that the Huff-BP-based multi-label classification model achieves relatively good classification results, and the digital signage audiences are mainly concentrated within the 4th Ring Road and near the 5th Ring Road.

A new backpropagation neural network classification model for prediction of incidence of malaria.

Malaria is an infectious disease caused by parasitic protozoans of the Plasmodium family. These parasites are transmitted by mosquitos which are common in certain parts of the world. Based on their specific climates, these regions have been classified as low and high risk regions using a backpropagation neural network (BPNN). However, this approach yielded low performance and stability necessitating development of a more robust model. We hypothesized that by spiking neuron models in simulating the characteristics of a neuron, which when embedded with a BPNN, could improve the performance for the assessment of malaria prone regions. To this end, we created an inter-spike interval (ISI)-based BPNN (ISI-BPNN) architecture that uses a single-pass spiking learning strategy and has a parallel structure that is useful for non-linear regression tasks. Existing malaria dataset comprised of 1296 records, that met these attributes, were used. ISI-BPNN showed superior performance, and a high accuracy. The benchmarking results showed reliability and stability and an improvement of 11.9% against a multilayer perceptron and 9.19% against integrate-and-fire neuron models. The ISI-BPNN model is well suited for deciphering the risk of acquiring malaria as well as other diseases in prone regions of the world.

An Approach to Segment and Track-Based Pedestrian Detection from Four-Layer Laser Scanner Data.

Pedestrian detection is a critical perception task for autonomous driving and intelligent vehicle, and it is challenging due to the potential variation of appearance and pose of human beings as well as the partial occlusion. In this paper, we present a novel pedestrian detection method via four-layer laser scanner. The proposed approach deals with the occlusion problem by fusing the segment classification results with past knowledge integration from tracking process. First, raw point cloud is segmented into the clusters of independent objects. Then, three types of features are proposed to capture the comprehensive cues, and 18 effective features are extracted with the combination of the univariate feature selection algorithm and feature correlation analysis process. Next, based on the segment classification at individual frame, the track classification is conducted further for consecutive frames using particle filter and probability data association filter. Experimental results demonstrate that both back-propagation neural network and Adaboost classifiers based on 18 selected features have their own advantages at the segment classification stage in terms of pedestrian detection performance and computation time, and the track classification procedure can improve the detection performance particularly for partially occluded pedestrians in comparison with the single segment classification procedure.

Analysis of DES Plaintext Recovery Based on BP Neural Network

Backpropagation neural network algorithms are one of the most widely used algorithms in the current neural network algorithm. It uses the output error rate to estimate the error rate of the direct front layer of the output layer, so that we can get the error rate of each layer through the layer-by-layer backpropagation. The purpose of this paper is to simulate the decryption process of DES with backpropagation algorithm. By inputting a large number of plaintext and ciphertext pairs, a neural network simulator for the decryption of the target cipher is constructed, and the ciphertext given is decrypted. In this paper, how to modify the backpropagation neural network classifier and apply it to the process of building the regression analysis model is introduced in detail. The experimental results show that the final result of restoring plaintext of the neural network model built in this paper is ideal, and the fitting rate is higher than 90% compared with the true plaintext.

Correlation-Based Ensemble Feature Selection Using Bioinspired Algorithms and Classification Using Backpropagation Neural Network.

A framework for clinical diagnosis which uses bioinspired algorithms for feature selection and gradient descendant backpropagation neural network for classification has been designed and implemented. The clinical data are subjected to data preprocessing, feature selection, and classification. Hot deck imputation has been used for handling missing values and min-max normalization is used for data transformation. Wrapper approach that employs bioinspired algorithms, namely, Differential Evolution, Lion Optimization, and Glowworm Swarm Optimization with accuracy of AdaBoostSVM classifier as fitness function has been used for feature selection. Each bioinspired algorithm selects a subset of features yielding three feature subsets. Correlation-based ensemble feature selection is performed to select the optimal features from the three feature subsets. The optimal features selected through correlation-based ensemble feature selection are used to train a gradient descendant backpropagation neural network. Ten-fold cross-validation technique has been used to train and test the performance of the classifier. Hepatitis dataset and Wisconsin Diagnostic Breast Cancer (WDBC) dataset from University of California Irvine (UCI) Machine Learning repository have been used to evaluate the classification accuracy. An accuracy of 98.47% is obtained for Wisconsin Diagnostic Breast Cancer dataset, and 95.51% is obtained for Hepatitis dataset. The proposed framework can be tailored to develop clinical decision-making systems for any health disorders to assist physicians in clinical diagnosis.

Communication modulation recognition algorithm based on STFT mechanism in combination with unsupervised feature-learning network

Aiming at the limitations of traditional communication modulation recognition algorithms, a novel recognition algorithm based on deep learning network far communication signal features is proposed in this paper. By introducing the two different computation mechanisms of STFT, two unsupervised feature-learning networks based on Restrict Boltzmann Machine (RBM) are respectively adopted for communication signal, where the network based on RBM is greatly improved in computing performance compared to the network based on convolutional Restrict Boltzmann Machine (CRBM), and greatly reduces the requirement for high-performance hardware in deep learning networks. In addition, as for signal modulation recognition and signal detection problems in communications, two modulation recognition networks are constructed by using the learning-to-STFT networks and the Back Propagation Neural Network (BPNN) classifier. Compared with the traditional modulation algorithms, our proposed algorithm in the paper can obtain better performance in the recognition accuracy, especially under the condition of low SNR.

A data transfer method for improving seed identification of maize (Zea mays) haploid breeding based on genetic similarity

AbstractMaize haploid breeding technology is able to identify haploid seeds non‐destructively, rapidly and at low cost with the help of Near‐infrared (NIR) spectral analysis. However, due to the hybridization of numerous parents and the low production rate of haploid, the haploid data collection becomes a burden for engineering this technology. Biologically, there are considerable similarities between the progeny of the same female parent and different male parents. Based on this advantage, similar spectral data can be transferred when the NIR technology is employed. A revised method of Transfer adaptive boost (TrAdaBoost) is proposed to improve identifying for the backpropagation neural network (BPNN) classifier. To avoid the negative transfer, a screening thresh is used to select out similar data, and the amount of these data are limited to join current training. The results show that the identification performances are improved significantly when the data amount is small. This method shows a high ability to make the seed identification more convenient for engineering maize haploid breeding.

Classification of EEG-based single-trial motor imagery tasks using a B-CSP method for BCI

Classifying single-trial electroencephalogram (EEG) based motor imagery (MI) tasks is extensively used to control brain-computer interface (BCI) applications, as a communication bridge between humans and computers. However, the low signal-to-noise ratio and individual differences of EEG can affect the classification results negatively. In this paper, we propose an improved common spatial pattern (B-CSP) method to extract features for alleviating these adverse effects. First, for different subjects, the method of Bhattacharyya distance is used to select the optimal frequency band of each electrode including strong event-related desynchronization (ERD) and event-related synchronization (ERS) patterns; then the signals of the optimal frequency band are decomposed into spatial patterns, and the features that can describe the maximum differences of two classes of MI are extracted from the EEG data. The proposed method is applied to the public data set and experimental data set to extract features which are input into a back propagation neural network (BPNN) classifier to classify single-trial MI EEG. Another two conventional feature extraction methods, original common spatial pattern (CSP) and autoregressive (AR), are used for comparison. An improved classification performance for both data sets (public data set: 91.25%±1.77% for left hand vs. foot and 84.50%±5.42% for left hand vs. right hand; experimental data set: 90.43%±4.26% for left hand vs. foot) verifies the advantages of the B-CSP method over conventional methods. The results demonstrate that our proposed B-CSP method can classify EEG-based MI tasks effectively, and this study provides practical and theoretical approaches to BCI applications.

Intelligent detection technology for leakage bag of baghouse based on distributed optical fiber sensor

In order to overcome the deficiency of traditional methods for damaged bag identification and location of baghouse in environmental protection industry, a real-time online identification and location system for leak bag of bag filter based on distributed optical fiber sensing technology of phase-sensitive optical time-domain reflectometer (Φ-OTDR) is proposed. Through the design of laying mode of optical fiber in filter bag of baghouse, the program flow of identifying and locating damaged filter bag is designed, and the real-time monitoring and positioning of filter bags is realized by distributed optical fiber sensing system. Wavelet packet decomposition (WPD) method is used to obtain the energy spectrum and energy entropy. Taking the energy ratio of the third sub-band frequency signal and the energy entropy as eigenvectors, the characteristic differences of optical fiber vibration signals between different types of good bags and different types of damaged bags with different leakage hole size and position are analyzed. The results show that the eigenvectors extracted by the method used in this paper can effectively describe the characteristics of the airflow vibration signals in the filter bags. Back propagation (BP) neural network algorithm is used to recognize the optical fiber vibration signals in damaged bags with different leakage hole sizes and positions. In addition, quantitative analysis method is used to identify the sample signals in different types of damaged bags. First, the airflow vibration signals in damaged bags with different leakage hole positions are identified while the leakage hole size remains the same. Second, the airflow vibration signals in leakage bags with different leakage hole sizes are identified while the leakage hole position remains the same. Finally, the BP neural network classifier is trained by the characteristics of the vibration signals in a filter bag, and it is used to identify the remaining field testing signals. Each type of vibration event is identified 10 times and the average recognition rate is calculated. The results show that the proposed classifier maintains high recognition stability and a high recognition rate for different types of damaged bags is obtained for the proposed method, which can reach higher than 90%.

Heuristic vs Metaheuristic Method: Improvement of Spoofed Fingerprint Identification in IoT Devices

The Internet of Thing (IoT) has become important due to the development of information technology that connects several aspects such as devices, services, and humans. Many IoT devices have been implemented the security system that uses biometrics sensor. The security system needs to secure the personal activity data recorded in IoT devices. Fingerprint is the one of common biometrics used in authentication but it has several security issues. Spoofed fingerprint is one of the risk security problems in biometrics authentication. This paper focuses on improving the security aspect in biometrics that is fundamental in IoT devices. Some optimization methods such as Nelder Mead, heuristic and Simulated Annealing and metaheuristic have been used to optimize the spoofed fingerprints identification with Back Propagation Neural Network classifier. Gray Level Co-occurrence Matrix is also considered as feature extraction of fingerprints in this research. The result of this study has shown that Nelder Mead method has a better improvement accuracy with a significant of two-tail p-value equal to 0.013. Moreover, Simulated Annealing method has been able to improve the accuracy in identification of spoofed fingerprint faster than Nelder Mead, but unsignificantly.