Nezhurina Marianna, Kuban State Technological University, Krasnodar, Russia
In the current paper we present a new Data Augmentation technique – Random Zeroing. Thistechnique is easy to implement, it can be usedalongside with the existing data augmentation techniques. We trained Convolutional Neural Networks on three datasets CIFAR10, MNIST and Fashion MNIST and compared validation accuracies of networks trained with different variations of Random Zeroing techniques.
Data Augmentation, Computer Vision, Convolutional Neural Networks, Random Erasing.
Lowri Williams, Eirini Anthi, Amir Javed, Pete Burnap, School of Computer Science & Informatics, Cardiff University, Cardiff, UK
The performance of emotive text classification using affective hierarchical schemes (e.g., WordNet-Affect) is often evaluated using the same traditional measures used to evaluate the performance of when a finite set of isolated classes are used. However, applying such measures means the full characteristics and structure of the emotive hierarchical scheme are not considered. Thus, the overall performance of emotive text classification using emotion hierarchical schemes is often inaccurately reported and may lead to ineffective information retrieval and decision making. This paper provides a comparative investigation of how four extended evaluation metrics which consider the characteristics of the hierarchical scheme can be applied and subsequently improve the performance of the classification of emotive texts. This study investigates the classification performance of three widely used classifiers, Naive Bayes, J48 Decision Tree, and SVM, following the application of the aforementioned methods. The results demonstrated that all methods improved the performance of all classifiers. However, the most notable improvement was recorded when a depth-based method was applied to both the testing and validation data, where the precision, recall, and F1-score were significantly improved by around 70 percentage points for each classifier.
Sentiment Analysis, Emotion Classification, Supervised Machine Learning, Hierarchical Classification, Natural Language Processing.
Shehab Eldeen Ayman1, Walid Hussein2, Omar H. Karam3, 1Department of Software Engineering, Faculty of Informatics and Computer Science, The British University in Egypt, Cairo, Egypt, 2Department of Computer Science, Faculty of Informatics and Computer Science, The British University in Egypt, Cairo, Egypt, 3Department of Information Systems, Faculty of Informatics and Computer Science, The British University in Egypt, Cairo, Egypt
Many real-time object recognition systems operate on two-dimensional images, degrading the influence of the involved objects third-dimensional (i.e., depth) information. The depth information of a captured scene provides a thorough understanding of an object in full-dimensional space. During the last decade, several region proposal techniques have been integrated into object detection. scenes’objects are then localized and classified but only in a two-dimensional space.Such techniques exist under the umbrella of two-dimensional object detection models such as YOLO and SSD. However, these techniques have the issue of being uncertain that an objects boundaries are properly specified intothescene. This paper proposes a unique region proposal and object detection strategy based on retrieving depth information forlocalization and segmentation of the scenes’ objects in real-time manner. The obtained results on different datasets show superior accuracy in comparison to the commonly implemented techniques with regards to not only detection but also apixel-by-pixel accurate localization of objects.
Real time object detection, region proposal, computer vision, RGBD object detection, two stage object detection.
Pablo Josué Francia Del Busto, Rodson Vladimir Ayme Tambra and Juan Antonio Flores Moroco, Department of Software Engineering, Universidad Peruana de Ciencias Aplicadas, Lima, Perú
Serverless models are one of the latest architecture models provided by Cloud vendors such as AWS and Microsoft, we are exploring serverless applications to develop a progressive web application that will help future developers and project leaders to better manage their projects. As seen in the investigationdone, we see Latin America being below the global average incompleting projects on time or within the planned budget. In this work we will explain the agile life cycle of the web application developed, focusing on the Scrum aspects of the tool design, the serverless architecture of the app and its following development, while also applying an analysis of the current projects to measurethe level of effectiveness that a proper Scrum method can haveon finishing a project correctly on time and within budget Journals.
Progressive app, Web application, FaaS, Serverless, Scrum, Agile.
Sameir A. Aziez, Electromechanical Engineering Dept., University of Technology, Dr. Asst. Prof. Saad M, khaleefah, Al- hikma college university, BAGHDAD – IRAQ, Bassam H. Abed, Electrical Engineering Dept., University of Technology, Iraq, Thamir R. Saeed, Electrical Engineering Dept., University of Technology, Iraq, Shaymaa A. Mohammed, Electrical Engineering Dept., University of Technology, Iraq, Ghufran. M. Hatema, Najaf Technical College, Al-Najaf Al-Ashraf, Iraq
The observation of dynamic systems is essential in many fields. Many algorithms are used to estimate this dynamic system. Short-Time Fourier transform (STFT) is one of these algorithms. This paper presents an optimized STFT to extract the Doppler properties of that system. The improvement reached 7-11% against the unoptimized process, and the processing speed was also affected by 35%.
Short-time Fourier transform, optimization, Doppler, Dynamic system.
Peter Hillmann, Diana Schnell, Harald Hagel, and Andreas Karcher, Department of Computer Science, Universität der Bundeswehr, Munich, Germany
The knowledge of the world is passed on through libraries. Accordingly, domain expertise and experiences should also be transferred within an enterprise by a knowledge base. Therefore, models are an established medium to describe good practices for complex systems, processes, and interconnections. However, there is no structured and detailed approach for a design of an enterprise model library. The objective of this work is the reference architecture of a repository for models with function of reuse. It includes the design of the data structure for filing, the processes for administration and possibilities for usage. A case study with industry demonstrates the practical benefits of reusing work already done. It provides an organization with systematic access to specifications, standards and guidelines. Thus, a further development is accelerated and supported in a structured way, the complexity remains controllable. The presented approach details various enterprise architecture frameworks. It provides benefits for development based on models.
Enterprise Architecture, Model Library, Business-IT-Alignment, Reference Architecture, Enterprise Repository for reusable Models.
Wei Liu1, Fang Wei Li2, Hai Bo Zhang3, Bo Li4, 1School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing, China, 2Chongqing Key Laboratory of Public Big Data Security Technology, Chongqing, China, 3School of Communication and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing, China, 4Chongqing Key Lab of Mobile Communications Technology，Chongqing Universityof Post and Communications
This paper studies a wireless power communication network(WPCN) based on orthogonal frequency division multiplexing (OFDM) with time reversal(TR). In this paper, the " Harvest Then Transmit " protocol is adopted, and the transmission time block is divided into three stages, the first stage is for power transmission, the second stage is for TR detection, and the third stage is for informationtransmission. The energy limited access point (AP) and the terminal node obtain energy fromthe radiofrequency signal sent by the power beacon (PB) to assist the terminal data transmission. The energy limited AP and the terminal node obtain energy from the radio frequency signal sent by the PB to assist the terminal data transmission. In the TR phase and the wireless information transmission (WIT) phase, the terminal transmits the TR detection signal to the AP using the collected energy, and the AP uses the collected energy to transmit independent signals to a plurality of terminals through OFDM. In order tomaximize the sum rate of WPCN, the energy collection time and AP power allocation are jointly optimized. Under the energy causal constraint, the subcarrier allocation, power allocation and time allocation of the whole process are studied, and because of the binary variables involved in the subcarrier allocation, the problem belongs to the mixed integer non-convex programming problem. the problem is transformed into a quasiconvex problem, and then binary search is used to obtain the optimal solution. The simulation results verify the ef ectiveness of this scheme. The results showthat the proposed scheme significantly improves the sum rate of the terminal compared to the reference scheme.
Wireless Powered Communication Network, Ttime Reversal, OFDM.
Muhammad E. H. Chowdhury, Tawsifur Rahman, Amith Khandakar, Sakib Mahmud, Department of Electrical Engineering, Qatar University, Doha, Qatar
The novelcoronavirus disease (COVID-19) is a highly contagious infectious disease. Even though there isa large pool of articles that showed the potential of using chest X-ray images in COVID-19 detection, a detailed study using a wide range of pre-trained convolutional neural network (CNN) encoders-based deep learning framework in screening viral, bacterial, and COVID-19 pneumonia are still missing. Deep learning network training is challenging without a properly annotated huge database. Transfer learning is a crucial technique for transferring knowledge from real-world objectclassification tasks to domain-specific tasks, and it may offer a viable answer. Although COVID-19 infection on the lungs and bacterial and viral pneumonia share many similarities, they are treated differently. Therefore, it is crucial to appropriately diagnose them. The authors have complied alarge X-ray dataset (QU-MLG-COV) consisting of 16,712 CXR images with 8851 normal, 3616 COVID-19, 1485 viral, and 2740 bacterial pneumonia CXR images. We employed image pre-processing methods and 21 deep pre-trained CNN encoders to extract features, which were then dimensionality reduced using principal component analysis (PCA) and classified into 4-classes. We trained and evaluated every cutting-edge pre-trained network to extract features to improve performance. CheXNet surpasses other networks for identifying COVID-19, Bacterial, Viral, and Normal, with an accuracy of 98.89 percent, 97.87 percent, 97.55 percent, and 99.09 percent, respectively. The deep layer network found significant overlaps between viral and bacterial images. The paper validates the network learning from the relevant area of the images by Score-CAM visualization.The performance of the various pre-trained networks is also thoroughly examined in the paper in terms of both inference time and well-known performance criteria.
Novel Coronavirus disease, COVID-19, viral pneumonia, bacterial pneumonia, deep learning, Convolutional neural network, Principal component analysis.
Muhammad E. H. Chowdhury1, Md Shafayet Hossain2, Sakib Mahmud1, Amith Khandakar1, 1Department of Electrical Engineering, Qatar University, Doha, 2713, Qatar, 2Department of Electrical, Electronic and Systems Engineering, Universiti Kebangsaan Malaysia, Bangi, 43600, Selangor, Malaysia
The adoption of dependable and robust techniques to remove electromyogram (EMG) artifacts from electroencephalogram (EEG) is essential to enable the exact identification of several neurological diseases. It is still a challenge to design an effective technique to eliminate EMG artifacts from EEG recordings, even though many classical signal processing-based techniques have been used in the past and only a few deep-learning-based models have been proposed very recently. In this work, deep learning (DL) techniques have been used to remove EMG artifacts from single-channel EEG data by employing four popular 1D convolutional neural networks (CNN) models for signal synthesis. To train, validate, and test four CNN models, a semi-synthetic publicly accessible EEG dataset known as EEGdenoiseNet has been used. The performance of 1D CNN models has been assessed by calculating the relative root mean squared error (RRMSE) in both the time and frequency domain and the temporal and spectral percentage reduction in EMG artifacts. Moreover, the average power ratios between five EEG bands to whole spectra are separately calculated. The U-Net model outperformed the other three 1D CNN models in most cases in removing EMG artifacts from EEG. U-Net achieved the highest temporal and spectral percentage reduction in EMG artifacts (90.01% and 95.49%); the closest average power ratio for theta, alpha, beta, and gamma band (0.55701, 0.12904, 0.07516, and 0.01822, respectively) compared to ground truth EEG (0.5429; 0.13225; 0.08214; 0.002146; and 0.02146, respectively). It is expected from the reported results that the proposed framework can be used for real-time EMG artifact reduction from multi-channel EEG data.
EEG, EMG artifacts, Deep Learning, Single Channel, Denoising, Convolutional neural network.
Lotfi Souifi1, Afef Mdhaffar1,2, Ismael Bouassida Rodriguez1, Mohamed Jmaiel1,2, and Bernd Freisleben3, 1University of Sfax, ENIS, ReDCAD Laboratory, B.P. 1173 Sfax, Tunisia, 2Digital Research Center of Sfax, 3021 Sfax, Tunisia, 3Dept. of Math. & Comp. Sci., Philipps-Universität Marburg, Germany
Due to the advancements of deep learning (DL), particularly in the areas of visual object detection and convolutional neural networks (CNN), insect detection in images has received a lot of attention from the research community in the last few years. This paper presents a systematic review of the literature on the topic of insect detection in images. It covers 50 research papers on the subject and responds to three research questions: i) type of dataset used; ii) detection technique used; iii) insect location. The paper also provides a summary of existing methods used for insect detection.
Systematic Literature Review (SLR), Deep Learning (DL), Object Detection, Insect Detection.
José Manuel Brotons Martínez, Economic and Financial Department, Miguel Hernández University, Elche, Spain
Since water is an essential element for agriculture, it is crucial to measure its productivity. In this regard, regions with a scarcity of water coexist with others that have an abundance of it, and whose cost is practically non-existent. So, to make the results comparable, we need to obtain a correct measurement, which will require setting a market price for water in areas where no price has yet been set. Therefore, the aim of this paper is to propose new productivity indicators based on fuzzy logic, whereby experts’ opinions about the possible price of the use of water as well as the annual variability of agricultural prices can be added. Therefore, the fuzzy willingness to pay (FWTP) and fuzzy willingness to accept (FWTA) methodology will be applied to create an artificial water market. The use of fuzzy logic will allow the uncertainty inherent in the experts’ answers to be collected. Ordered Weighted Averaging (OWA) operators and their different extensions will allow different aggregations based on the sentiment or interests reflected by the experts. These same aggregators, applied to the prices of the products at origin, will make it possible to create new indicators of the economic productivity of water. Finally, through an empirical application for a pepper crop in south-eastern Spain we can visualize the importance of the different indicators and their influence on the final results.
Water Economic Productivity, OWA, Fuzzy Willingness to Pay, Fuzzy Willingness to Accept.
Behruz Saidov and Vladimir Telezhkin, South Ural State University (National Research University), Chelyabinsk, Russia
This article explores the methods of information processing in automated control systems based on ultrasonic transceivers. Modern automated control systems (ACS), including those for special purposes, widely use information processing methods that use digital technologies and network data exchange between various sensors. Recently, the ultrasonic sensor has been widely used in various applications, in particular, in transmitting and receiving information. The advantage of these systems (ultrasonic sensor) is, on the one hand, the possibility of communication, both with close and remote access, on the other hand, a high probability of leak detection and elimination of transmitted information. Ultrasonic transducers and ultrasonic sensors are devices that generate or receive ultrasonic energy. They can be divided into three broad categories: transmitters, receivers and transceivers. Transmitters convert electrical signals to ultrasound, receivers convert ultrasound to electrical signals, and transceivers can both transmit and receive ultrasound. The purpose of this work is to process information in automated control systems using an ultrasonic sensor. To solve this problem, an experiment was conducted to study transceivers at different distances. According to the results of the pilot study, it was concluded that it is possible to transmit and receive an ultrasonic sensor effectively at a frequency of 20 kHz at a distance of about 10 meters.
information processing, ultrasonic signal, automated control systems.