Multi-Lingual Language Variety Identification using Conventional Deep Learning and Transfer Learning Approaches
Abstract: Language variety identification tends to identify lexical and semantic variations in different varieties of a single language. Language variety identification helps build the linguistic profile of an author from written text which can be used for cyber forensics and marketing purposes. Investigating previous efforts for language variety identification, we hardly find any study that experiments with transfer learning approaches and/or performs a thorough comparison of different deep learning approaches on a range of benchmark datasets. So, to bridge this gap, we propose transfer learning approaches for language variety identification tasks and perform an extensive comparison of them with deep learning approaches on multiple varieties of four widely spoken languages, i.e., Arabic, English, Portuguese, and Spanish. This research has treated this task as a binary classification problem (Portuguese) and multi-class classification problem (Arabic, English, and Spanish). We applied two transfer learning Bidirectional Encoder Representations from Transformers (BERT), Universal Language Model Fine-tuning (ULMFiT), three deep learning-Convolutional Neural Networks (CNN), Bidirectional Long Short Term Memory (Bi-LSTM), Gated Recurrent Units (GRU), and an ensemble approach for identifying different varieties. A thorough comparison between the approaches suggests that the transfer learning based ULMFiT model outperforms all other approaches and produces the best accuracy results for binary and multi-class language variety identification tasks.
Keywords: Language variety identification, deep learning, transfer learning, binary classification.
Received July 25, 2021; accepted December 13, 2021
A Novel Binary Search Tree Method to Find an Item Using Scaling
Praveen Pappula
School of Computer Science and Artificial Intelligence, SR University, India
This email address is being protected from spambots. You need JavaScript enabled to view it.
Abstract: This Approach comprises of methods to produce novel and efficient methods to implement search of data objects in various applications. It is based on the best match search to implement proximity or best match search over complex or more than one data source. In particular with the availability of very large numeric data set in the present day scenario. The proposed approach which is based on the Arithmetic measures or distance measures called as the predominant Mean based algorithm. It is implemented on the longest common prefix of data object that shows how it can be used to generate various clusters through combining or grouping of data, as it takes O(log n) computational time. And further the approach is based on the process of measuring the distance which is suitable for a hierarchy tree property for proving the classification is needed one for storing or accessing or retrieving the information as required. The results obtained illustrates overall error detection rates in generating the clusters and searching the key value for Denial of Service (DOS) attack 5.15%, Probe attack 3.87%, U2R attack 8.11% and R2L attack 11.14%. as these error detection rates denotes that our proposed algorithm generates less error rates than existing linkage methods.
Keywords: Clustering, classification, KNN, vector quantization, mean based search, scaling.
Received May 1, 2020; accepted July 4, 2021
Transfer Learning for Feature Dimensionality Reduction
Abstract: Transfer learning is a machine learning methodology by which a model developed for achieving a task is exploited for another related job. Many pre-trained image classification models trained on ImageNet are used for transfer learning. These pre-trained networks could also be used for classifying out of domain images by retraining them. This paper, along with the existing application for these pre-trained models, is also being exploited for feature dimensionality reduction. Many dimensionality reduction methods are available; the pre-trained image models will help us perform both image feature extraction and dimensionality reduction in a single go using the same network. The fine-tuning of the fully connected layers of the pre-trained network is done to extract the image features; along with this fine-tuning, some more tweaking is done on the fully connected layers of these models to reduce the image feature dimensionality. Here, VGG-16 and VGG-19 are the pre-trained models considered for feature vector generation and dimensionality reduction. An analysis of the efficiency of features generated by these pre-trained networks in classifying the out-of-domain images is done. Three different variants of VGG-16 and VGG-19 are analysed. All the three variants developed gave an AUC value above 0.8, which is considered good.
Keywords: Dimensionality reduction, fine-tuning, transfer learning, VGG-16, VGG-19.
Received August 10, 2020; accepted October 13, 2021
A Fusion Approach Based on HOG and Adaboost Algorithm for Face Detection under Low- Resolution Images
|
Farhad Navabifar* Department of Electrical Engineering, Mobarakeh branch, Islamic Azad University, Isfahan, Iran This email address is being protected from spambots. You need JavaScript enabled to view it. |
Mehran Emadi Department of Electrical Engineering, Mobarakeh branch, Islamic Azad University, Isfahan, Iran This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: Detecting human faces in low-resolution images is more difficult than high quality images because people appear smaller and facial features are not as clear as high resolution face images. Furthermore, the regions of interest are often impoverished or blurred due to the large distance between the camera and the objects which can decrease detection rate and increase false alarms. As a result, the performance of face detection (detection rate and the number of false positives) in low-resolution images can affect directly subsequent applications such as face recognition or face tracking. In this paper, a novel method, based on cascade Adaboost and Histogram of Oriented Gradients (HOG), is proposed to improve face detection performance in low resolution images, while most of researches have been done and tested on high quality images. The focus of this work is to improve the performance of face detection by increasing the detection rate and at the same time decreasing the number of false alarms. The concept behind the proposed combination is based on the a-priori rejection of false positives for a more accurate detection. In other words in order to increase human face detection performance, the first stage (cascade Adaboost) removes the majority of the false alarms while keeping the detection rate high, however many false alarms still exist in the final output. To remove existing false alarms, a stage (HOG+SVM) is added to the first stage to act as a verification module for more accurate detection. The method has been extensively tested on the Carnegie Melon University (CMU) database and the low-resolution images database. The results show better performance compared with existing techniques.
Keywords: Face detection, cascade adaboost, histogram of oriented gradients, low-resolution image.
Received November 20, 2020; accepted January 10, 2022
Enhanced-AODV Routing Protocol to Improve Route Stability of MANETs
|
Ako Abdullah Department of Computer Science University of Sulaimani, Iraq This email address is being protected from spambots. You need JavaScript enabled to view it. |
Emre Ozen Department of Information Technology Eastern Mediterranean University (EMU), Mersin 10 Turkey This email address is being protected from spambots. You need JavaScript enabled to view it. |
Husnu Bayramoglu Department of Information Technology Eastern Mediterranean University (EMU), Mersin 10 Turkey This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: Establishing and maintaining link stability in Mobile Ad hoc Networks (MANETs) is one of the key challenging issues. Topology changes in MANET because overhead traffic that leads to consuming extra energy of nodes as well as decreasing the performance of routing protocols. Thus, a comprise approach should be considered during the design of a routing scheme in MANETs to deal with challenges incurred by the mobility of the nodes. In this study, a simple efficient routing scheme called Enhanced_AODV (E-AODV) is proposed, aiming to enhance Ad Hoc On-Demand Distance Vector (AODV) routing protocol performance by constructing the most stable and reliable route from source to the destination node. In this routing scheme, the remaining lifetime of links and hop count are the metrics considered for calculating the Route Stability Factor (RSF) that can be utilized as a cost metric to establish the best route between source and destination node. The simulation results reveal that the proposed E-AODV routing scheme effectively outperforms the conventional AODV routing protocol and Stable and Bandwidth Aware Dynamic Routing Protocol (SBADR) in terms of packet delivery ratio, average network throughput, average end-to-end delay, and normalized routing overhead.
Keywords: MANET, routing protocol, shortest path, route stability, network density.
Received December 24, 2020; accepted February 13, 2022
Face Anti-Spoofing System using Motion and Similarity Feature Elimination under Spoof Attacks
|
Aditya Bakshi Department of Computer Science and Engineering, Shri Mata Vaishno Devi University, India This email address is being protected from spambots. You need JavaScript enabled to view it. |
Sunanda Gupta Department of Computer Science and Engineering, Shri Mata Vaishno Devi University, India This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: From border control to mobile device unlocking applications, the practical utility of biometric system can be seriously compromised due face spoofing attacks. So, face recognition systems require greater attention to combating face spoofing attacks. As, face spoofing attacks can be easily propelled through 3D masks, video replays, and printed photos so we are presented face recognition system using motion and similarity features elimination under spoof attacks against the Replay Attack and Institute of Automation, Chinese Academy of Sciences (CASIA) databases. In this paper a calculative analysis has been done by firstly segmenting the foreground and background regions from the input video using Gaussion Mixture Model and secondly by extracting features i.e., face, eye, and nose and applied 26 image quality assessment parameters on spoof face videos under different illumination lighting conditions. The results attained using Replay Attack and CASIA databases are extremely competitive in discriminating from fake traits with paralleled viz-a-viz other approaches. Different machine learning classifiers and their comparative analysis with existing approaches has been shown.
Keywords: Linear discriminant analysis, support vector, gaussian mixture model, image quality assessment.
Received January 13, 2021; accepted December 9, 2021
Secure of Transmission Systems in the Visible Range of Light with the Power Line Communication Interface
Abstract: The paper discusses the issue of creating a secure environment for information transmission in an association dependent on Noticeable Light Correspondence (VLC) innovation utilizing a Power Line Correspondence (PLC) modem. Throughout the examination, an investigation of homegrown and unfamiliar writing and patent documentation was done, which affirmed the significance of this point and the need to improve and adjust the innovation for homegrown associations, the conceivably better security of the framework from unapproved admittance to information was validated in correlation with wired and Wi-Fi organizations., which executes information transmission dependent on VLC innovation, utilizing Light-Emitting Diode (LED). Lighting sources as transmitters of the correspondence framework, with the coordination of the PLC interface. The utilization of the PLC interface makes it conceivable to improve on the establishment and execution of the VLC communicating modules since the data signal is provided to the last through the electrical cables that give capacity to the lighting installations. An evaluation of the working states of an information transmission framework dependent on VLC innovation with a PLC interface was completed, which uncovered that with the base admissible sign to-clamor proportion equivalent to 6 dB, the channel data transfer capacity is 8 Mbps, and the bit mistake rate will in general zero. The examination results can be utilized to assemble a corporate organization utilizing VLC innovation with a PLC interface, and discover their application for additional investigation of this innovation.
Keywords: VLC, Li-Fi, PLC, FSO, data visible light, passing module, optical signal telecommunication system.
Received June 11, 2021; accepted August 15, 2021
Crop Disease Prediction with Convolution Neural Network (CNN) Augmented With Cellular Automata
|
Kiran Sree Pokkuluri Department of Computer Science and Engineering, Shri Vishnu Engineering College for Women (A), Bhimavaram This email address is being protected from spambots. You need JavaScript enabled to view it. |
SSSN Usha Devi Nedunuri Department of Computer Science and Engineering, University College of Engineering-Jawaharlal Nehru Technological University, Kakinada This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: Food security is the primary concern of any country, and crop diseases are the major threats to this. Each stage of the crop will be affected by various diseases starting from seeding to ripeness. The spread of the crop diseases is very rapid, and identification of this is challenging as the infrastructure is very less to monitor the same. After a thorough literature survey, we understood there are several ways of predicting the disease and yield prediction. We have developed two new and robust classifiers, one which processes images to predict the crop's diseases, and the second one uses the weather data to predict the same. Both classifiers use deep-learning technique Convolution Neural Networks (CNN) augmented with six neighborhood cellular automata to predict the crop disease and yield. This work will be first of its kind to develop two classifiers for six crop disease prediction. The average time to compute the yield of a particular crop is less than 0.5 nanoseconds. The first classifier is named as CNN-CA-I, which was trained/tested to process 245 different crop species and 132 diseases associated with these crops where image segmentation is done with higher accuracy, thus strengthening the disease recognition system. We gave collected public datasets of 12, 45,678 images diseases and leaves of healthy plants taken in ideal conditions. This model reports an accuracy of 92.6% on a tested standard dataset for disease and yield prediction. The second classifier is CNN-CA-W that predicts crop disease trained and tested with environment data.8,52.624 datasets are collected from ECMWF for processing the weather data to predict the crop's condition and thus reporting the yield of the crop. This model reports an accuracy of 90.1% on a tested standard dataset.
Keywords: CNN, cellular automata, disease prediction, image segmentation, weather prediction.
Received May 6, 2020; accepted November 24, 2020
Compatibility Themed Solution of the Vehicle Routing Problem on the Heterogeneous Fleet
|
Metin Bilgin Department of Computer Engineering, Bursa Uludağ University, Turkey This email address is being protected from spambots. You need JavaScript enabled to view it. |
Nisanur Bulut Department of Computer Engineering, Bursa Uludağ University, Turkey This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: In this study, we discuss the solution to the vehicle routing problem for a heterogeneous fleet with a depot and a time window satisfied by meeting customer demands with various constraints. A 3-stage hierarchical method consisting of transportation, routing, and linear correction steps is proposed for the solution. In the first stage, customer demands have the shortest routing. They were clustered using the annealing simulation algorithm and assigned vehicles of appropriate type and equipment. In the second stage, a genetic algorithm was used to find the optimal solution that satisfies both the requirements of the transported goods and the customer requirements. In the third stage, an attempt was made to increase the optimality by linear correction of the optimal solution found in the second stage. The unique feature of the application is the variety of constraints addressed by the problem and the close proximity to real logistics practice.
Keywords: Time window, vehicle routing problem, multiple traveling salesmen problem, heterogeneous fleet, simulated annealing algorithm, genetic algorithm, optimization.
Received May 19, 2020; accepted September 22, 2021
Deep Learning Based Hand Wrist Segmentation using Mask R-CNN
|
GokulaKrishnan Elumalai School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, India This email address is being protected from spambots. You need JavaScript enabled to view it. |
Malathi Ganesan School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, India This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: Deep learning is one of the trending technologies in computer vision to identify and classify objects. Deep learning is a subset of Machine Learning and Artificial Intelligence. Detecting and classifying the object was a challenging task in traditional computer vision techniques, and now there are numerous deep learning techniques scaled up to achieve this. The primary purpose of the research is to detect and segment the human hand wrist region using deep learning methods. This research is widespread to deep learning enthusiasts who needs to segment custom objects using instance segmentation. We demonstrated a segmented hand wrist using the Mask Regional Convolutional Neural Network (R-CNN) technique with an average accuracy of 99.73%. This work also compares the performance evaluation of baseline and custom Hand Wrist Mask R-CNN. The achieved validation class loss is 0.00866 training and 0.02736 validation; both the values are comparatively deficient compared with baseline Mask R-CNN.
Keywords: Hand wrist, segmentation, mask R-CNN, fast R-CNN, faster R-CNN, object detection.
Received June 27, 2020; accepted October 13, 2021
AVL Based Settlement Algorithm and Reservation System for Smart Parking Systems in IoT-based Smart Cities
|
Hikmet Canli Department of Computer Engineering, Duzce University, Turkey This email address is being protected from spambots. You need JavaScript enabled to view it. |
Sinan Toklu Department of Computer Engineering, Gazi University, Turkey This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: In Internet of Things (IOT)-based smart cities, negative reasons such as cost, energy and air pollution when searching for a parking space increase the importance of smart parking systems. In this study, a two-stage hybrid approach is proposed so that drivers can find a parking space that will consume the least time and energy. The first stage focuses on car parks having at least one free parking space located near the target address in an n diameter circumference, which are also open for business. An AVL tree-based hierarchical structure is created with driving time from the starting point to each car park and walking time from each car park to the destination, and it focuses on the most appropriate car park. In the second stage, the most suitable parking space is searched and made available, if found, in hierarchical parking monitoring system. In order to demonstrate the effectiveness of the approach, the results compared with hierarchical, hierarchical Binary Search Tree (BST) and non-hierarchical solutions in terms of energy and time performance are shown on a simulation. Proposed approach gave the best result with 99% energy efficiency. In addition, a dynamic cloud-based reservation system was proposed for the parking lot determined in the study.
Keywords: AVL tree, adjacency lists, smart cities, smart cities, cloud, internet of things.
Received November 1, 2020; accepted November 24, 2021
A Novel Genetic Algorithm with db4 Lifting for Optimal Sensor Node Placements
|
Ganesan Thangavel* Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, India This email address is being protected from spambots. You need JavaScript enabled to view it. |
Pothuraju Rajarajeswari Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, India This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: Target coverage algorithms have considerable attention for monitoring the target point by dividing sensor nodes into cover groups, with each sensor cover group containing the target points. When the number of sensors is restricted, optimal sensor node placement becomes a key task. By placing sensors in the ideal position, the quality of maximum target coverage and node connectivity can be increased. In this paper, a novel genetic algorithm based on the 2-D discrete Daubechies 4 (db4) lifting wavelet transform is proposed for determining the optimal sensor position. Initially, the genetic algorithm identifies the population-based sensor location and 2-D discrete db4 lifting adjusts the sensor location into an optimal position where each sensor can cover a maximum number of targets that are connected to another sensor. To demonstrate that the suggested model outperforms the existing method, A series of experiments are carried out using various situations to achieve maximum target point coverage, node interconnectivity, and network lifetime with a limited number of sensor nodes.
Keywords: Wireless sensor network, target point coverage, node connectivity, sensor deployment, genetic algorithm, two-dimensional db4 lifting, network lifetime.
Received March 19, 2021; accepted October 21, 2021
Separable High Capacity Reversible Data Hiding Algorithm for Encrypted Images
|
Iyad Jafar Computer Engineering Department, The University of Jordan, Jordan This email address is being protected from spambots. You need JavaScript enabled to view it. |
Khalid Darabkh Computer Engineering Department, The University of Jordan, Jordan This email address is being protected from spambots. You need JavaScript enabled to view it. |
Fahed Jubair Computer Engineering Department, The University of Jordan, Jordan This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: This paper presents a separable Reversible Data Hiding Algorithm for Encrypted Images (RDHEI) that consists of three phases. The encryption phase in the algorithm circularly shifts the columns in the image by a random amount, blocks the image into equal and regular blocks and maps them to irregular blocks generated based on Hilbert filling curve, and finally complements a random subset of the blocks. The embedding phase is essentially an adapted version of the modification of the prediction errors algorithm that is applied to each block in the encrypted image independently. In the decryption phase, and since the algorithm is separable, the user can extract the data only, decrypt the image only, or can perform both actions depending on the type of keys he has. When compared to a very similar and recent algorithm, performance evaluation proved the ability of the proposed algorithm in increasing the embedding capacity with reasonable quality of the directly decrypted image. In terms of the security, the analytical and quantitative assessment showed the superiority of the proposed algorithm in protecting the encrypted image.
Keywords: Encryption, embedding capacity, privacy, reversible data hiding.
Received April 23, 2021, accepted December 2, 2021
A Lightweight Hybrid Intrusion Detection Framework using Machine Learning for Edge-Based IIoT Security
Abstract: Due to the development of cloud computing and Internet of Things (IoT) environments, such as healthcare systems, telecommunications and Industry 4.0 or Industrial IoT (IIoT) many daily services are transformed. Therefore, Security issues become useful to better protect these novel technologies. IIoT security represents a real challenge for industry actors and academic research. A set of security approaches, such as intrusion detection are integrated to improve IIoT environments security. Hence, an Intrusion Detection System (IDS) aims to monitor, detect an intrusion in real time and then make reliable decisions. Many recent IDS incorporate Machine Learning (ML) techniques to improve their Accuracy (ACC), precision and Detection Rate (DR). This paper presents a hybrid IDS for Edge-Based IIoT Security using ML techniques. This new hybrid framework is based on misuse and anomaly detection using K-Nearest Neighbor (K-NN) and Principal Component Analysis (PCA) techniques. Specifically, the K-NN classifier has been incorporated to improve detection accuracy and make effective decision and the PCA is used for an enhanced feature engineering and training process. The obtained results have proven that our proposed Framework presents many advantages compared with other recent models. It gives good results with 99.10% ACC, 98.4% DR 2.7% False Alarm Rate (FAR) on NSL-KDD dataset and 98.2% ACC, 97.6% DR, 2.9% FAR on Bot-IoT dataset.
Keywords: IoT security, edge-based IIoT, intrusion detection, ML, K-NN, PCA. NSL-K, Bot-IoT.
Received August 6, 2021, accepted December 9, 2021
Environmental Noise Adaptable Hearing Aid using Deep Learning
|
Soha A. Nossier Department of Biomedical Engineering, Medical Research Institute, Alexandria University, Egypt This email address is being protected from spambots. You need JavaScript enabled to view it. |
M. R. M. Rizk Department of Electrical Engineering, Faculty of Engineering, Alexandria University, Egypt This email address is being protected from spambots. You need JavaScript enabled to view it. |
|
Saleh El Shehaby Department of Biomedical Engineering, Medical Research Institute, Alexandria University, Egypt This email address is being protected from spambots. You need JavaScript enabled to view it. |
Nancy Diaa Moussa Department of Biomedical Engineering, Medical Research Institute, Alexandria University, Egypt This email address is being protected from spambots. You need JavaScript enabled to view it. |
Abstract: Speech de-nosing is one of the essential processes done inside hearing aids, and has recently shown a great improvement when applied using deep learning. However, when performing the speech de-noising for hearing aids, adding noise frequency classification stage is of a great importance, because of the different hearing loss types. Patients who suffer from sensorineural hearing loss have lower ability to hear specific range of frequencies over the others, so treating all the noise environments similarly will result in unsatisfying performance. In this paper, the idea of environmental adaptable hearing aid will be introduced. A hearing aid that can be programmed to multiply the background noise by a weight based on its frequency and importance, to match the case and needs of each patient. Furthermore, a more generalized Deep Neural Network (DNN) for speech enhancement will be presented, by training the network on a diversity of languages, instead of only the target language. The results show that the learning process of DNN for speech enhancement is more efficient when training the network using diversity of languages. Moreover, the idea of adaptable hearing aid is shown to be promising and achieved 70% overall accuracy. This accuracy can be improved using a larger environmental noise dataset.
Keywords: Adaptable hearing aid, MFCC, neural networks, noise classification, speech enhancement.
Received December 1, 2020; accepted December 9, 2021