Deep Belief Neural Network Research Articles

Moth Monarch Optimization-Based Deep Belief Network in Deception Detection System

Deception is the action of causing a person to believe something, which is known to be lying with the provision of evidence to support such false beliefs with certain intensions. Identification of the deceptive characteristics manually is a challenging problem for the researchers. Thus, an automatic deception detector is necessary to be developed in order to ensure higher accuracy. Accordingly, this paper proposes a novel deception detector method called Moth Monarch optimization-based Deep Belief Neural Network (MMO-DBN). The proposed MMO-DBN classifier undergoes the phases of feature extraction and classification. Initially, the input speech signals are pre-processed to remove the noise present in the signal and subjected to feature extraction to extract the significant features, such as Mel Frequency Cepstral Coefficients (MFCC), Spectral Kurtosis, Spectral Spread, Spectral Centroid, minimum blood pressure, maximum blood pressure, respiration rate, and Tonal Power Ratio. Then, these extracted features are subjected to classification using Deep Belief Neural Network (DBN), which is trained with the proposed Moth Monarch optimization (MMO) algorithm that is the integration of Monarch Butterfly Optimization (MBO) and Moth Search (MS) algorithm. The performance of the proposed MMO-DBN is analyzed using the metrics, namely accuracy, sensitivity, and specificity. The proposed method obtained the higher accuracy, sensitivity, and specificity of 0.984, 0.9836, and 0.9375, respectively that shows the superiority of the proposed MMO-DBN in deception detection.

Deep learning strategies for foetal electrocardiogram signal synthesis

One of the most difficult tasks for the physicians is to acquire a quality foetal electrocardiogram (fECG) to analyze, manage and plan according to the condition of the foetus in the womb. Hence the foetal electrocardiogram signal is not preferred to execute the analysis to monitor the Foetal condition. Other traditional methods are being used to access the foetal condition. The foetal electrocardiogram signal can be acquired either by using invasive or non- invasive techniques. Since the invasive technique is harmful for the foetus, non-invasive technique is mostly adopted. The foetal electrocardiogram signal can be acquired only after twenty five weeks the foetus is developed in the womb, which is referred as the Antepartum period. This article portrays the use of Deep learning techniques for non-invasive foetal electrocardiogram signal synthesis using artificial intelligent techniques. Convolutional neural network (CNN), Deep belief neural networks (BNN) and Back propagation Neural Network (BPNN) have been utilized and tested for the proposal. The outcomes and performance are compared with reference to the synthesized high quality foetal electrocardiogram signal.

Data‐driven nonlinear chemical process fault diagnosis based on hierarchical representation learning

AbstractRepresentation extraction is crucial in data‐driven process monitoring, and deep neural network (DNN) is an efficient tool for extracting representations from considerable process data. This study proposes a hierarchical representation learning (HRL) method that integrates the deep belief neural (DBN) network and support vector data description (SVDD) for efficient nonlinear chemical process fault diagnosis. First, hierarchical representations containing meaningful process information are generated through a DBN network by utilizing generally massive normal operating process data. Second, an SVDD‐based decision‐making system is constructed using generally small‐sized faulty data. Three experimental studies are then conducted. A comparison of results with those of several state‐of‐the‐art methods reveal the suitability of the HRL method for process monitoring due to its two main advantages. First, DNN has a superior representative ability and generates representations with richer process information than conventional data‐driven methods. Second, the HRL method utilizes available process data and is suitable for practical conditions in which considerable normal operating data but limited small‐sized faulty data are available.

Analyzing patient health information based on IoT sensor with AI for improving patient assistance in the future direction

• Traditional health care system fails to predict the exact patient health information. • For the above issue, an IoT sensor with AI is used to predict exact patient details. • Healthcare mobile application is used to collect the patient’s information easily. • The gathered patient details are processed according to the IGDBN. Internet of Things (IoT) and Artificial Intelligence (AI) play a vital role in the upcoming years to improve the assistance systems. The IoT devices utilize several sensor devices that able to collect a large volume of data in different domains which is processed by AI techniques to make the decision about the assistance problems. Among several applications, in this work, IoT with AI is used to examine the healthcare sectors to improve patient assistance and patient care in the future direction. Traditional health care assistance system fails to predict the exact patient health information and needs which reduces the accuracy of patient assistance process. For these issues, an IoT sensor with AI is used to predict the exact patient details such as fitness tracker, medical reports, health activity, body mass, temperature, and other health care information which helps to choose the right assistance process. Healthcare mobile application is used to achieve this goal and collect the patient’s information. This information is shared in the cloud environment, which is accessed and processed by applying the optimized machine learning techniques. The gathered patient details are processed according to the iterative golden section optimized deep belief neural network (IGDBN). The introduced network examines the patient’s details from the previous health information which helps to predict the exact patient health condition in the future direction. The efficiency of IoT sensor with an AI-based health assistance prediction process is developed using MATLAB tool. Excellence is determined in terms of precision (99.87), loss error (0.045), simple matching coefficient (99.71%), Matthews correlation coefficient (99.10%) and accuracy (99.86%).

Sparse-FCM and deep learning for effective classification of land area in multi-spectral satellite images

Remote sensing plays a major role in crop classification, land use classification, and land cover classification such that the information for the classification is assured with the help of the satellite images. This paper concentrates on the land use classification and proposes an optimization algorithm, called Firefly Harmony Search (FHS) for training the Deep Belief Neural Network (DBN). The FHS algorithm is the integration of the Firefly Algorithm and Harmony Search Algorithm (HSA), which tunes the weights of DBN to perform the multi-class classification. For the effective classification, the multispectral image is subjected to the sparse Fuzzy C-Means to form segments such that the feature extraction is effective, free from dimensionality issues and computational complexities. The features extracted from the segments of the multi-spectral images include vegetation indices and statistical features. Then, these features are fed to the DBN, which is tuned using the FHS algorithm for performing the land use classification. Experimentation using four datasets proves the effectiveness of the proposed multi-class classification approach. The accuracy, sensitivity, and specificity of the method are found to be 0.9317, 0.9568, and 0.0379, respectively, that is effective over the existing land use classification methods.

RETRACTED ARTICLE: A Robust Face Emotion Recognition Approach through Optimized SIFT Features and Adaptive Deep Belief Neural Network

We, the Editor and Publisher of Journal of Applied Security Research, have retracted the following article: Yenumaladoddi Jayasimha & R. Venkata Siva Reddy, A Robust Face Emotion Recognition Approach through Optimized SIFT Features and Adaptive Deep Belief Neural Network, Journal of Applied Security Research, 2019, DOI: 10.1080/19361610.2019.1656471. This article contains extensive and continuous content similarity across all sections with the previously published article: Yenumaladoddi Jayasimha & R. Venkata Siva Reddy (2019) A robust face emotion recognition approach through optimized SIFT features and adaptive deep belief neural network, Intelligent Design Technologies, 13.3, 379–390. DOI: 10.3233/IDT-190022. The substantial similarity between the articles is not acknowledged. Following a request by the corresponding author to withdraw this article from Journal of Applied Security research, we discovered the previously published article. We contacted the corresponding author multiple times but have received no response. Therefore, we have determined that retracting this article is the best course of action. We have been informed in our decision-making by our policy on publishing ethics and integrity and the COPE guidelines on retractions. The retracted article will remain online to maintain the scholarly record, but it will be digitally watermarked on each page as “Retracted.”

ELECTROENCEPHALOGRAM CLASSIFICATION USING VARIOUS ARTIFICIAL NEURAL NETWORKS

Electroencephalography (EEG) is the oldest diagnostic tool used in the field of neurosciences. When a neurologist sees an EEG report, he can point out important neural defects in a person but many a times diagnoses were missed and it is just impossible for human brain to process all the data in EEG. Nowadays many deep machine learning architectures are developed to understand the information contained in EEG signals. This study reviews various literatures on classification of EEG signals using various artificial neural networks like Convoluted neural network, Recurrent Neural Network, Deep Belief neural networks and hybrids.

Fuzzy DDBN: Fuzzy Dragon Deep Belief Neural Network and interesting features points for activity recognition

Activity recognition is the interest gaining research area, as the need for monitoring and controlling the public and the society to ensure the detection of the suspects and the illegal activities is of prime importance. The process of recognizing the activity of the humans is employed in various applications, mainly in the field of security to identify and detect the suspects. Accordingly, this paper uses a novel method named as the Fuzzy-DDBN classifier to categorize the human activities in the video. At first, the keyframes are extracted from the video based on the Bhattacharya similarity measure, and the keyframes are subjected to feature extraction using the Scale Invariant Feature Transform (SIFT) and the Spatio-Temporal Interest (STI) descriptor. The features extracted from the descriptors are fed to the classifier for classification that in turn, uses the GMM clustering. The classified output from the proposed Fuzzy-DDBN classifier, which is the combination of the fuzzy and the Dragon Deep Belief Neural Network (DDBN) classifier is merged using the correlation coefficients. The proposed method is experimented using two standard datasets to prove the superiority of the method with an accuracy of 0.98, specificity at a rate of 0.981, and sensitivity at a rate of 0.98 respectively.

ACO–MKFCM: An Optimized Object Detection and Tracking Using DNN and Gravitational Search Algorithm

Object tracking is a dynamic optimization process based on the temporal information related to the previous frames. Proposing a method with higher precision in complex environments is a challenge for researchers in the field of study. In this research, efficient object detection and movement tracking video is proposed. Here we are considering the input video sequence is PETS and Hall monitor video. Kernel fuzzy c-means procedure still endures several downsides, such as reduced convergence rate, obtaining stuck in the local minima as well as at risk to initialization level of sensitivity. Ant colony optimization algorithm is a new population-based optimization technique that has actually been utilized effectively for resolving numerous complicated difficulties. This research suggested a new strategy called ACO–MKFCM to resolve MKFCM initialization issue making use of ant colony optimization (ACO) algorithm to locate optimum first cluster centres for the MKFCM, hence enhance all applications associated fuzzy clustering such as foreground segmentation in image processing. Initially, the background and foreground separation is done by hybridization of modified kernel fuzzy c means algorithm (MKFCM) with ant colony optimization. The recommended new technique is intellectual and also vibrant clustering method for splitting up of non-static object. Then object detection and tracking is done by gravitational search algorithm based deep belief neural network. The implementation will be in MATLAB. The performance of the suggested technique is appraised by means of precision, recall, F-measure, FPR, FNR, PWC, FAR, similarity, specificity, and accuracy. From the empirical effects, the future work outperforms than the state of art work. Here the proposed method attains maximum precision and recall value for both PETS and Hall monitor video when analyzed to the existent algorithm.

Improved Intrusion Detection System with Optimization Enabled Deep Neural Networks

Cyber-crimes are prevailing at the extreme in the today’s technical world as the massive usage of the internet is on the peak among the world users, raising the security and privacy concerns. Thus, the paper concentrates on the intrusion detection mechanism in the networks, which is performed using the optimization-based deep belief neural networks (DBN). Input data is classified using the DBN classifier and the complexity associated with the classification is relieved through the feature selection strategy for which the Bhattacharya distance is employed. The DBN training is performed using Levenberg–Marquardt (LM) algorithm and Bird Swarm Algorithm (BSA), which is decided based on the minimal mean square error. The intrusion detection affords the security and privacy to the data. The analysis of the methods is presented using the KDD cup dataset and the comparative analysis is performed based on the accuracy, sensitivity, and specificity. The accuracy, sensitivity, and specificity of the BSA-DBN approach of intrusion detection are found to be 96.45%, 94.07%, and 96%, respectively.

Deep Learning-based Framework for Smart Sustainable Cities

Pure air is vital for sustaining human life. Air pollution causes long-term effects on people. There is an urgent need for protecting people from its profound effects. In general, people are unaware of the levels to which they are exposed to air pollutants. Vehicles, burning various kinds of waste, and industrial gases are the top three onset agents of air pollution. Of these three top agents, human beings are exposed frequently to the pollutants due to motor vehicles. To aid in protecting people from vehicular air pollutants, this article proposes a framework that utilizes deep learning models. The framework utilizes a deep belief network to predict the levels of air pollutants along the paths people travel and also a comparison with the predictions made by a feed forward neural network and an extreme learning machine. When evaluating the deep belief neural network for the case study undertaken, a deep belief network was able to achieve a higher index of agreement and lower RMSE values.

Classification and Detection of Faults in Induction Motor using Dwt with Deep Learning Methods under the Time-Varying and Constant Load Conditions

This article proposed a method to detect the faults in multi-winding induction motor using Discrete Wavelet transform combined with Deep Belief Neural Network (DBNN). This technique relies on the instantaneous reactive power signal decomposition, from which detail coefficients and wavelet approximations are extracted which are termed as features. In order to obtain a robust diagnosis, this article proposed to classify the feature vectors extracted from DWT analysis of power signals using DBNN (Deep Belief Neural Network) to distinguish the motors state. Subsequently, in order to validate the proposed approach, a three phase squirrel cage induction machine is simulated under MATLAB software. To check the effectiveness of the proposed method of fault diagnosis the motor is simulated in different simulation environments like time varying load and constant load condition. Promising results were obtained and the performance of DBNN i.e. 99.75% accuracy. The proposed algorithm is compared with various other state-of-art methods and the comparison proves its robustness in diagnosing the fault in motors.

Iris Recognition from an Image at Lengthy Distance by using Deep Belief Neural Network (DBN)

Now a days, biometric is a one of the best method which is used for the detection of person is iris recognition. A large portion of different frameworks are equally introduces for individual ID like as distinguishing proof cards or tokens, mystery codes, passwords, and so on. Yet, the issues of these sort of frameworks are, the mystery codes and passwords can be split, the recognizable proof cards can be harmed. Subsequently the successful strategy for the individual recognizable proof is vital. Iris gives the unmistakable data about an individual. Iris recognition is the process of identifying persons automatically using their iris. Iris provides the distinctive information about a person. This paper exhibits the deep learning-based methodology for the iris acknowledgment. Firstly, the picture is pre-handled to get the precise area of the iris. From that point onward, iris locale is extricated utilizing Hough Transform, which is pursued with the division and standardization of the iris area utilizing the Daugman's Rubber sheet model. When the division is played out, the features are separated by utilizing the Local Gradient Pattern (LGP) and ScaT-LOOP that is the mixture of Scattering transforms (ST), Tetrolet transforms (TT), and Local Optimal Oriented Pattern (LOOP) descriptors. At last, steepest slope based Deep Belief Neural Network (DBN) is used for the iris acknowledgment. The exhibition of iris acknowledgment utilizing the DBN classifier is assessed regarding precision, False Rejection Rate (FRR) and False Acceptance Rate (FAR). The proposed iris acknowledgment strategy accomplishes the most extreme precision of 97.96%, negligible FAR of 0.493%, and insignificant FRR of 0.48% that shows its predominance.

A robust face emotion recognition approach through optimized SIFT features and adaptive deep belief neural network

Facial expression is a significant indications for nonverbal communication between individuals. The assignment of face emotion recognition is predominantly intricate for two reasons. Initial one is...

Characterization of $S_1$ and $S_2$ Heart Sounds Using Stacked Autoencoder and Convolutional Neural Network

This paper proposes a new technique for the identification of fundamental heart sounds (HSs), namely, $S_{1}$ and $S_{2}$ from the cardiac cycle, the first and foremost step in the automated HSs analysis for the detection of pathological events, without incorporating time-interval informations between $S_{1}$ and $S_{2}$ or electrocardiogram signal as reference. The motive of this paper is to demonstrate that the reliable $S_{1}$ and $S_{2}$ classification performances based on the combinatory feature (CF) derived from higher order moments and cepstral-based domain can still be achieved, under the circumstances where the timing interval information might not be easily understood due to cardiac abnormalities. Using deep neural networks approach, a stacked autoencoder (SAE) based on the CF is proposed for the classification of fundamental HSs. Experiments are conducted on both publicly available and recorded HSs signals for the validation of the proposed method. The SAE using the proposed CF achieves better classification results in recognizing $S_{1}$ and $S_{2}$ in comparison to well-known classifiers such as deep belief neural network, support vector machine, Naive Bayes, linear discriminant analysis, and boosting ensemble. The proposed method shows higher classification rate in terms of accuracy, sensitivity, and specificity by considering CF, which uses Mel-frequency cepstral coefficients and its derivative features. A second approach for addressing the problem of $S_{1}$ and $S_{2}$ identifications is carried out by employing 1-D convolutional neural network that uses the signals directly to learn the relevant features by its own for the recognition.

Utilizing IoT wearable medical device for heart disease prediction using higher order Boltzmann model: A classification approach

Globally, the prognosis of heart disease can be improved by early diagnosis and treatment. However, existing automatic systems for diagnosing heart disease are hampered by the requisite big data. This paper introduces an Internet of Things-based medical device for collecting patients’ heart details before and after heart disease. The information, which is continuously transmitted to the health care center, is processed using the higher order Boltzmann deep belief neural network (HOBDBNN). The deep learning method learns heart disease features from past analysis, and achieves efficiency by the effective manipulation of complex data. Following experiments, the performance of the system is evaluated based on characteristics such as f-measure, sensitivity, specificity, loss function, and receiver operating characteristic (ROC) curve. The HOBDBNN method and IoT-based analysis recognize heart disease with 99.03% accuracy with minimum time complexity (8.5 s), effectively minimizing heart disease mortality by reducing the complexity of diagnosing heart disease.

Fuzzy dragon deep belief neural network for activity recognition using hierarchical skeleton features

In computer vision, human activity recognition is an active research area for different contexts, such as human–computer interaction, healthcare, military applications, and security surveillance. Activity recognition is performed to recognize the goals and actions of one or more people from a sequence of observations based on the actions and the environmental conditions. Still, there are numbers of challenges and issues, which motivate the development of new activity recognition method to enhance the accuracy under more realistic conditions. This paper proposes an error-based fuzzy dragon deep belief network (error-based DDBN), which is the integration of fuzzy with DDBN classifier, to recognize the human activity from a complex and diverse scenario, for which the keyframes are generated based on the Bhattacharya coefficient from the set of frames of the given video. The key frames from the Bhattacharya are extracted using the scale invariant feature transform, color histogram of the spatio-temporal interest dominant points, and hierarchical skeleton. Finally, the features are fed to the classifier, where the classification is done using the proposed error-based fuzzy DDBN to recognize the person. The experimentation is performed using two datasets, namely KTH and Weizmann for analyzing the performance of the proposed classifier. The experimental results reveal that the proposed classifier performs the activity recognition in a better way by obtaining the maximum accuracy of 1, a sensitivity of 0.99, and the specificity of 0.991.

Deep Neural Architectures for Highly Imbalanced Data in Bioinformatics.

In the postgenome era, many problems in bioinformatics have arisen due to the generation of large amounts of imbalanced data. In particular, the computational classification of precursor microRNA (pre-miRNA) involves a high imbalance in the classes. For this task, a classifier is trained to identify RNA sequences having the highest chance of being miRNA precursors. The big issue is that well-known pre-miRNAs are usually just a few in comparison to the hundreds of thousands of candidate sequences in a genome, which results in highly imbalanced data. This imbalance has a strong influence on most standard classifiers and, if not properly addressed, the classifier is not able to work properly in a real-life scenario. This work provides a comparative assessment of recent deep neural architectures for dealing with the large imbalanced data issue in the classification of pre-miRNAs. We present and analyze recent architectures in a benchmark framework with genomes of animals and plants, with increasing imbalance ratios up to 1:2000. We also propose a new graphical way for comparing classifiers performance in the context of high-class imbalance. The comparative results obtained show that, at a very high imbalance, deep belief neural networks can provide the best performance.

Applying a Neural Network Ensemble to Intrusion Detection

Abstract An intrusion detection system (IDS) is an important feature to employ in order to protect a system against network attacks. An IDS monitors the activity within a network of connected computers as to analyze the activity of intrusive patterns. In the event of an ‘attack’, the system has to respond appropriately. Different machine learning techniques have been applied in the past. These techniques fall either into the clustering or the classification category. In this paper, the classification method is used whereby a neural network ensemble method is employed to classify the different types of attacks. The neural network ensemble method consists of an autoencoder, a deep belief neural network, a deep neural network, and an extreme learning machine. The data used for the investigation is the NSL-KDD data set. In particular, the detection rate and false alarm rate among other measures (confusion matrix, classification accuracy, and AUC) of the implemented neural network ensemble are evaluated.

Low-order Nonlinear Finite-Impulse Response Soft Sensors for Ionic Electroactive Actuators Based on Deep Learning

This paper introduces a soft sensor (SS) for the estimation of the deflection of a polymeric mechanical actuator. The actuator is based on ionic polymer-metal composites (IPMCs). Applications of IPMCs have been proposed in fields such as robotics, surgery, and aerospace, to mention the most interesting ones. In such application fields, both the complexity and the size of the actuating system are of chief importance. An SS can be, therefore, preferred to hardware measuring the actuator output, for estimating the actuator motion. Also, low-order models are of interest to limit the computational load, which can be a constraint in real-time applications. To this aim, several data-driven nonlinear finite-impulse response (NFIR) models have been investigated. Data, used for the model identification, have been acquired, in controlled environmental conditions, by using swept signals as the input to the IPMC actuator. Linear and nonlinear models, based on principal component analysis, shallow, and deep neural networks (NNs), have been investigated, for different model orders. The best results have been obtained by an SS based on a fifth-order NFIR model, implemented by a deep belief NN.