International Journal of Intelligent Engineering and Systems

12

Authors: Arnaud Nanfak, Charles Hubert Kom, Samuel Eke

As a major piece of equipment in the electrical power transmission and distribution, the rapid and accurate assessment of emerging or existing internal faults in power transformers is a key factor in the safe and stable operation of the power grid. This paper proposes a hybrid approach to fault diagnosis for these transformers. This approach is based on ensemble bagged tree classification and training subsets obtain by a conventional pre-processing method. Two pre-processing approaches are performed, the first based on the maximum concentrations of the dissolved gas samples and the second based on the minimum concentrations of the dissolved gas samples. For each training subset, an ensemble classifier is constructed with as inputs the Rogers ratios, Gouda ratios, dissolved gas concentrations and their associations. The proposed hybrid methods are established with 475 samples of training dataset, tested on 117 samples dissolved gas analysis (DGA) data and validated on International Electrotechnical Commission (IEC) TC10 database. The performances of the proposed diagnostic models are evaluated and a comparison is done compared with the following diagnostic methods: IEC ratios method (IRM), Rogers ratios method (RRM), three ratios technique (TRT), Gouda’s triangle (GT), and self-organizing map (SOM) clusters. The results found by computer simulations carried out by the matrice laboratory (MATLAB) software show that, of the two pre-processing approaches used, the one based on the minimum sample concentration gives better results than the one based on the maximum concentration. In terms of fault type, the best diagnostic model using the minimum concentration-based pre-processing approach has a diagnostic accuracy of 94.02%, compared to 92.31% for the best diagnostic model using the maximum concentration-based pre-processing approach. This is lower than the 97.25% for SOM clusters and 96.58% for GT but higher than the 59.83% for RRM, 81.19% for IRM and 93.16% for TRT. In terms of fault severity, the best diagnostic model using the minimum concentration-based pre-processing approach has a diagnostic accuracy of 81.20%, compared to 74.36% for the best diagnostic model using the minimum concentration-based pre-processing approach. This result is lower than the 83.76% for TRT and 89.74% for GT but higher than the 49.57% for RRM, 66.67% for IRM and 78.90% for SOM clusters.

35

Authors: Billel Arbaoui, Yusraw O-manee, Patimoh Yuso, Yuttachai Hareebin

Self-awareness emotional intelligence symptoms will be unavoidable in the job. These symptoms are generally regarded as acceptable by certain standards, as they are an important part of keeping people informed (or commonly known as situational awareness). However, the most recent research focuses on external components of situational awareness aspects and lacks adaptable methods to deal with the individual's situational awareness dynamics. This study investigated the use of an agent-based modeling strategy for situational awareness in the workplace. It focuses on environmental and personal factors that influence the level of situational awareness at work in a dynamic way. The outcomes included a range of scenarios that corresponded to various personality traits and environmental conditions. Finally, equilibria analysis and automated logical verification were employed to evaluate this computational model in order to check whether there were any conceptual defects, as indicated in the literature. The suggested computational agent-based model has shown rational behavior patterns that are consistent with existing psychology literature on situational awareness in the workplace.

57

Authors: Manar Bashar Mortatha, Samer Salah Thabit, Hawraa R. Abdul Ameer, Riyadh Rahef Nuiaa

During the image enhancement process, hidden information, poor capturing device quality lead to poor image contrast, insufficient user experience, and an inappropriate data collecting environment setting have all been noted as serious concerns. Histogram equalization techniques have been used to solve the challenges described above. Nonetheless, the images obtained using these methods are frequently impacted by unwanted artifacts, and unnatural appearances effects. Due to that, this research presented a novel strategy for enhancing contrast called Dynamic Clip Limit Window Size Histogram Equalization. The proposed technique uses a new fitness function of combining the discrete entropy and root mean square error parameters. The proposed strategy's qualitative and quantitative result is validated and assessed against 5 state-of-the-art strategies (HSQHE, ACLTSHE, CLAHE, AEIHE, and IAECHE). The proposed technique proved its high capability to produce the best average DE, CII, and SSI values for the Faces-199 dataset (7.869, 1.000, and 0.999, respectively).

81

Authors: Mohammed Shakir Oudah, Abeer Tariq Maolood

With the rapid use of the Internet, and the increase in the numbers of smart and mobile devices, the internet of things (IoT) fields had become more crucial fields in the interest of developers. Many aspects related to authentication and privacy communication during an interaction between remote users and IoT devices should be enabled and achieved in a lightweight and secure manner. Due to the elliptic curve digital signature (ECDSA)’s features that are secure and lightweight compared with other public key algorithms; this paper presents a new IoT authentication model that incorporates a modified ECDSA. But unfortunately, the familiar use of ECDSA in Blockchain showed some problems related to the capability of revealing the random private integer, which leads to private key disclosure, and hence funds theft, this paper highlights this problem and proposes modifications to the ECDSA to make it more reliable. In addition, the proposed model combines a modified ECDSA and Shamir’s secret sharing (SSS). The combination can give better results in establishing a more securely authentication agreement and robustness to resolve the standard algorithm attenuation, this modification involved a variant of ECDSA’s calculations in signature processing, and employing Shamir’s Secret Sharing to further protect the random private integer. Also, the proposed model achieves lower overhead communications by splitting and redistributing authentication calculations roles between the system’s entities. After security analysis, the proposed algorithm exhibits ability to resist potential threats, also, a comparison with other related works demonstrates that the proposed algorithm performs fewer arithmetic operations, relatively, by decreasing the number of modular inverse operations, so this feature lowers resources-efforts and high performance, moreover, the theoretical analysis indicates that the proposed algorithm efficiently manages the scalability of the system.

103

Authors: Abbas Salman Hameed, Dheyaa T. Al-Zuhairi

This paper proposes a peak to average power ratio (PAPR) reduction companding scheme for single carrier frequency division multiple access. The proposed technique has multi-level compression (MLC) stages where a signal is adjusted according to its instant value with respect to these levels. To validate MLC function, PAPR, bit error rate (BER), average power (P_av), and power spectral density (PSD) are evaluated throw simulation. MLC improved PAPR significantly with low BER degradation. Using 16-QAM, MLC enhances PAPR by 4.38 dB while only 1.6 dB additional SNR is required to maintain 10^(-3) BER. MLC performance is compared with the recent companding techniques; Mu-law, Exponential, Trapezoidal, and tangent rooting. The BER simulation results show the superiority of MLC over these techniques. For PAPR reduction, different from other techniques, both Mu-law and MLC reduce PAPR significantly with close values. Finally, the Exponential and MLC preserve P_av with less side lobes in PSD.

128

Authors: Ganapathy Kavitha, Kandasamy Vanitha

Psoriasis diagnosis from gene expression data requires efficient machine learning classifiers with the finest pre-processing methods to overcome the outlier and missing value problems. This paper developed relative entropy fuzzy score (REFS) normalization and genetic weighted k-nearest neighbors (GWKNN) imputation for enhancing pre-processing and hybrid multi-kernel universum support vector machines (MKUSVM) - radial basis function extreme learning machine (RBFELM) classifier for accurate psoriasis recognition. REFS normalization utilizes improved fuzzification and relative entropy in the gene fuzzy score estimation to reduce the skewness and outlier gene data. GWKNN imputation performs symbolic regression of weighted k-nearest neighbors (KNN) with genetic programming (GP) for missing value estimation. Finally, the hybrid MKUSVM-RBFELM classifier classifies the psoriasis gene data with features selected using mutual-information parameter. Evaluated over public dataset GSE55201, the proposed model achieved 91.67 % accuracy, 92.58 % precision, 90.94 % recall, 91.75 % F-measure, and 0.8452 MCC values with a reduced processing time of 0.373 seconds.

151

Authors: Deepak Jayaprakash Doggalli, Sunil Kumar B S

One of the primary causes of cancer-related mortality is liver cancer. Computed tomography (CT) is a commonly utilised imaging technique for the assessment and staging of hepatic tumors. Manually scanning volumetric CT scans is time intensive and ambiguous. Despite the fact that numerous deep learning models for semantic segmentation have been developed, the U-Net model remains extremely successful. In this work, we propose a method for segmenting liver tumors from abdominal CT images that is entirely based on the U-Net model and demonstrate the model's simplicity and efficacy. The U-Net architecture is employed at two levels. The first level of segmentation is used to segment the liver from CT slices, while the second level is utilised to segment tumours from masked CT images. Using CT scans from the LiTS dataset, the proposed technique achieved a dice global score of 94 percent for liver segmentation and 73 percent for tumour segmentation.

172

Authors: S. Aarthi, M. Shanmugasundari

This research provides a two-class single transmit queuing model. We also calculate the model’s execution proportions under a vague environment. The main purpose of this inquiry is to compare the results of a single transmit queuing model based on fuzzy queuing theory and intuitionistic fuzzy queuing theory. Triangular fuzzy numbers and triangular intuitionistic fuzzy numbers are used to describe the entry (arrival) and administration (service) rates. The fuzzy queuing theory model’s evaluation metrics are supplied as a range of values, but the intuitionistic fuzzy queuing theory model offers a wide range of values. An analysis is offered to discover quality measures utilizing a proposed methodology in which the fuzzy values are retained as is without being altered into crisp values, hence we can use the proposed method to draw scientific conclusions in an uncertain environment. Two numerical problems are solved to demonstrate the sustainability of the suggested method. Subsequently, prototype components were exposed to sensitivity analyses. Sensitivity testing is used to find discrepancies between the two groups when calculating their execution proportions. We employed the triangular fuzzy number in an intuitionistic fuzzy environment in this study, accounting for the degree of comfort and refusal so that the sum of both values is always less than 1. For this type of fuzzy integer, we gave various non-normal arithmetic procedures. The proposed formulations are simple and direct, having been devised using classical algebraic mathematics. This strategy is simple and straightforward to use in actual situations. The nearest interval number is then used to round a TIFN. The key benefit of this approach is that using a multi-section algorithm, we can quickly solve a bound unbridled optimization problem with coefficients as TIFN. The present methodologies and strategies are intended to be applicable to a variety of contemporary decision-making challenges in areas of economic share, finance, administration, and ecology, which are the focus of our future study.

194

Authors: Pallavi Hallappanavar Basavaraja, Shanmugarathinam Ganesarathinam

Osteoporosis is a condition wherein bone tissue deteriorates and bone strength deteriorates. Over time, the disease that can lead bone to become more permeable and fragile makes it more prone to breaking. Early recognition of osteoporosis via an X-ray approach will enhance diagnosis and treatment practices, but it will also assist in preventing national economic loss via mass screening and awareness. A novel early osteoporosis diagnosis model is developed for X-ray images in this research work. The following four primary steps are used to construct a unique Osteoporosis detection model in this study: “(a) pre-processing, (b) feature extraction, (c) optimal feature selection, and (d) Osteoporosis detection”. Gabor filtering (noise reduction) and histogram equalization are used to pre-process the obtained raw data (X-ray) (quality enhancement). Features such as “Active shape model (ASM), active appearance model (AAM), gray level co-occurrence matrix (GLCM), mean local gradient pattern (M-LGP), and improved median ternary pattern (I-MTP)” are recovered from the pre-processed data. Following that, a new hybrid optimization model chooses the best features from the retrieved features. The cat guided hummingbird foraging algorithm (CGHFA) is a conceptual combination of the basic artificial hummingbird algorithm (AHM) and the cat hunting optimization algorithm (CHOA). The deep learning classifiers in the Osteoporosis detection phase are trained using these ideally selected characteristics. The newly created ensemble-of-classifiers model is used to represent the osteoporosis diagnostic phase. “Quantum deep neural network (QDNN), improved deep convolution neural network (I-DCNN) and recurrent neural network (RNN)” are some of the deep learning classifiers that is employed here. All of these classifiers are trained using the optimal features available. The loss function of DCNN is improved via harmonic mean based cross-entropy function. The final detection performance will be calculated by combining the results obtained from all of these characteristics (by taking the mean). Finally, the effectiveness of the anticipated model is validated by a comparative examination. Accordingly, the detection accuracy attained by the proposed deep ensemble model +CGHFA at Learn_rate=60 is 90.7%, at Learn_rate=70 is 92.14%, at Learn_rate=80 is 93.482% and at Learn_rate=90 is 94.8%, which is higher than the existing models.

219

Authors: Loai Alamro, Yuhanis Yusof, Nooraini Yusoff

Human head attribute classification (HHAC) is a fundamental and substantial research domain in pattern recognition and computer vision. However, recent HHAC networks do not consider the correlation of common characteristics among the attributes over different regions of the human head. To address the above problem, this study proposes a multi-output convolutional neural network to jointly learn the features of human head attributes with common characteristics. The proposed network contains two convolutional blocks and five output layers, where each output layer learns to predict a specific group of human head attributes. In order to properly learn the correlation among the human head attributes, this study divides these attributes into five groups: hair, face, style, accessories, and appearance. Extensive experiments showed that the proposed network obtained an average classification accuracy of 95.29% and 97.93% on the challenging CelebA and LFWA datasets, respectively. Thus, the proposed network is approximately 2% and 10% superior to the closest competitor (i.e., PS-MCNN) on both datasets. In addition, the proposed network achieved higher classification accuracy compared to the existing networks almost in all human head attributes. That findings demonstrate the effectiveness of the proposed network and the attributes grouping method in learning the correlations among human head attributes correctly.

244

Authors: Noor Jabbar Abraham, Hazim G. Daway, Rafid Abbas Ali

Enhancement of images with low light has become an important role in the field of digital image processing, especially when you take the image at low or irregular lighting levels. In this study, a new algorithm was proposed to enhance images with low light based on the development of the Dark Chanel Prior (DCP), Where the image was first improved using this DCP then the lightness component in (Lab) space was improved using the sigmoid mapping. The proposed method was compared with several algorithms by using non-reference quality measures as naturalness image quality evaluator and image quality evaluator using LIME data. By analyzing the results we note the success of the proposed method in improving images with low light, with obtaining the best contrast and lightness compared to the rest of the methods, where the proposed method obtained the best quality values of natural image quality evaluator (3.462) and perception image quality evaluator (35.714).

261

Authors: Abdullah Fawzi Shafeeq, Inaam Ibrahim Ali

This study proposed a simple method for determining the best position and size of distributed generation (DG) and static VAR compensator (SVC) in the radial distribution system (RDS), as well as the impact of employing two different types of DG on total losses. The target functions are designed to reduce the system’s overall power losses, improve the voltage stability margin (VSM), and reduce drop voltages. This method relies on the deployment of many sizes of DGs or SVCs in each bus and comparing their impacts of them on the system’s overall losses and taking optimal sizing and placement of them if that given minimum losses. Moreover, the proposed strategy does not require extra work to set the parameter settings and weighting factors to obtain the objective functions. To assess the performance of the suggested approach, five instances were evaluated during DG and SVC installation in this study. The suggested approach is tested on the RDS’s IEEE 33_bus and 69_bus to ensure its practicality. This method decreased losses by 92.72 % for IEEE 33_bus and 96.91 % for IEEE 69_bus. In the 33_bus, the minimum VSM is improved from 0.6674 to 0.9697 p.u, and the minimum bus voltage is improved from 0.9038 to 0.994 p.u. While in the 69_bus, the minimum VSM is enhanced from 0.6835 to 0.9761 p.u and the minimum bus voltage is enhanced from 0.9093 to 0.9958 p.u, with the results compared to other existing methodologies.

284

Authors: Anindita Septiarini, Rizqi Saputra, Andi Tedjawati1 Masna Wati, Hamdani Hamdani

Sarong is a traditional cloth typically worn during formal or religious events which conventionally woven using the traditional loom. Samarinda is one of Indonesia’s regions with a sarong with a distinctive pattern. Nevertheless, the majority of indigenous people cannot distinguish the various motifs of Samarinda sarongs from those of other regions in Indonesia (non-Samarinda). Therefore, it is necessary to classify the motif of sarongs. This work proposed a pattern recognition method based on computer vision for sarongs motif classification. This method required adequate features to achieve the optimal results. Accordingly, the appropriate color and texture features were investigated to obtain the most discriminative ones. This work generated features using color moments, Gray Level Co-occurrence Matrix (GLCM), and Local Binary Pattern (LBP). The most discriminatory features were selected using the Correlation Based Feature Selection (CFS) and then fed into Artificial Neural Network (ANN). A total of 1000 images were used to evaluate the method and achieved the highest performance with an accuracy value of 100%.

309

Authors: Noor Abdul Khaleq Zghair, Ahmed Sabah Al-Araji

The primary goal of a mobile robot is to reach the desired goal by traversing an optimized path defined according to some criteria such as time, distance, and safety of the robot from any obstacles that may be in its way. Therefore, the backbone of autonomous mobile robots (AMRs) is path planning and avoiding obstacles. Many algorithms for path planning and obstacle avoidance have been presented by many researchers and each of these algorithms has several benefits and drawbacks. This paper focuses on comparing the performance of several metaheuristic algorithms that result in a more efficient, smoother, and shorter path for the mobile robot to reach the target in a complex environment. These algorithms include particle swarm optimization (PSO), chaotic particle swarm optimization (CPSO), modified chaotic particle swarm optimization (MCPSO), and firefly algorithm (FA). On the other hand, the paper proposes a hybrid algorithm by combining the FA and the MCPSO, namely the (HFAMCPSO). To demonstrate the effectiveness of the proposed algorithm in terms of the optimum cost function and obtaining the shortest path length, the optimal solution is compared to those of other path planning algorithms. Moreover, inverse dynamic and kinematic modeling are utilized to obtain the best torque and the best velocity actions for the wheels of the autonomous mobile robot. The proposed hybrid (FAMCPSO) algorithm provides enhancement on the path length equals (43.3%) and (25.5%) compared to the radial cell decomposition (RCD) and the A* algorithm, respectively. Moreover, the enhancement on the path length equals (2.3%) and (22.7%) compared to the firefly algorithm (FA) and the genetic algorithm (GA), respectively. All methods are simulated in an environment with static obstacles using the 2018b MATLAB package.

336

Authors: Sara Ahmed, Fatma A. Omara

Cloud computing has gained many attentions. Workflow scheduling one of the most important issues in cloud computing. It involves mapping tasks onto cloud resources – Virtual machines (VMs), to improve scheduling performance. Because the existing heterogeneous earliest finish time (HEFT) algorithm is considered one of the best algorithm, so the work in this paper propose a new algorithm based on HEFT algorithm; called modified heterogeneous earliest finish time (M-HEFT); to reduce the tradeoff among make span, resource utilization, and load balance. The proposed M-HEFT consists of two phases; task prioritization and task-VM mapping. In Task prioritization phase, a priority will be provided to each task in directed acyclic graph (DAG) as in the original HEFT algorithm. According to task-VM phase, tasks allocate to resources according to length of tasks and the load of available VMs with considering load balance. To evaluate the performance of the proposed algorithm, a comparative study has been done among the proposed algorithm and three existed algorithms. The experimental results show that the proposed algorithm outperforms the other algorithms by minimizing make span by 29%, improve resource utilization by 53% and load balance by 18% in average.

365

Authors: Prakit Intachai, Peerapol Yuvapoositanon

In this paper, the performance and robustness in remaining useful life (RUL) estimation of turbofan engines of the singular spectrum analysis and long short term memory (SSA-LSTM) algorithm with the window length optimization are investigated. Specifically, we are interested in how the optimized SSA-LSTM algorithm, i.e., the SSA-LSTM algorithm with its window length being optimized, responds to the variation in the number of cells and in the number of hidden layers of the LSTM structure. The performance and robustness of an SSA-LSTM algorithm are evaluated by two metrics, namely the root mean square (RMSE) ratio and the sensitivity ratio. Four optimized SSA-LSTM algorithms and two non-optimized SSA-LSTM algorithms were tested with the prototype dataset from the turbofan engines datasets. The simulation results showed that all the optimized SSA-LSTM algorithms performed much better in terms of both the RMSE ratios and the sensitivity ratios than the non-optimized SSA-LSTM algorithms. The RMSE ratios of the optimized SSA-LSTM algorithms with the window length of L=T/2,L=T/4,〖L=log⁡(T)〗^2.3 and L=24 were 38.737%, 33.967%, 22.896% and 11.780% respectively. In the same order, the sensitivity ratios of all the optimized algorithms were 2.067%, 1.495%, 1.555% and 1.444% respectively. Among all the algorithms being evaluated, the L=24 optimized SSA-LSTM algorithm provided the best performance in terms of both the RMSE ratio and the sensitivity ratio and hence confirming its best performance and robustness for RUL estimation.

390

Authors: Kittapon Chatwongtong, Tosaphol Ratniyomchai

In general, the train has consumed energy from the traction substation via the conductor during acceleration or motoring mode. This mode consumes most of the energy. However, during deceleration or braking mode, energy can be returned to the power terminal by regenerative braking management. This paper presents the energy loss reduction in the conductor transmission system and energy distribution mitigation from the traction substations by using the on-board supercapacitor, which is a storage device for storing regenerative braking energy. The results include the first case without a supercapacitor and the second case with supercapacitor operated on piece-wise linear state of charge (SOC) control. They are obtained by the single-train simulation on MATLAB and the power flow calculation by current injection method (CIM) on the peak time of the DC Mass Rapid Transit (MRT) Purple Line in Bangkok, Thailand. This research can effectively save energy distribution, energy loss and energy waste by approximately 3.5 %, 10 % and 20 % respectively.

416

Authors: Hind Zeyada, Sayed Abdo Mohamed, Gamal Tharwat, Hany Harb, Ayman Nasr Hamed

The interference of reflected waves from multiple elementary scatterers produces speckle, which appears as a granular noise in synthetic aperture radar (SAR) images. These speckles in SAR images cause difficulty in image interpretation, which reduces the effectiveness of image segmentation and classification. In this paper, we propose an effective solution using generative adversarial networks (GAN) to decrease speckle noise while preserving texture features.The convolutional block attention module (CBAM) boosts the essential features in upsampled data and supports the creation of a final denoised real InSAR data.The proposed deep-learning method proved that CBAM enhanced the peak signal-to-noise ratio using the GAN. Furthermore, CBAM improved structural similarity (SSIM) to 99 and achieved the minimum mean squared error. The despeckle performance was enhanced using GAN-ResUNet in which the SSIM equals 0.999. The denoising performance proved that the use of GAN with CBAM as the generator inspired by ResUNeT achieved the best performance compared to other experiments results.

439

Authors: Abdul Azeez, Suraiya Tarannum

To meet the increasing demand for wireless applications requires efficient performance in orthogonal frequency division multiplexing (OFDM) and multi-input multi-output (MIMO). The existing methods involve applying various optimization techniques to increase the performance of the partial transmit sequence (PTS) and OFDM system to reduce the peak average to power ratio (PAPR). This research proposes prime learning ant lion optimization (PL-ALO) method to reduce the PAPR and bit error rate (BER) of the system. The tournament selection technique is applied to increase the exploitation related to the best fitness agent that improves the PL-ALO model performance. The tournament selection technique randomly performs tournaments among the best fitness search agent in the system. The square-root raised cosine (SRC) precoding and Mu law companding techniques were applied to improve the efficiency of the system. The comparison between discrete cosine transform (DCT) and fast fourier transform (FFT) techniques was carried out in the performance analysis. The DCT-based method has higher efficiency than FFT-based techniques in the system. The PL-ALO method has higher efficiency than existing techniques of particle swarm optimization (PSO)- grey wolf optimization (GWO) and multi-objective mayfly algorithm (MOMA). The average PAPR of selective mapping (SLM) is 8.243, PTS-ant lion optimization (ALO) is 3.962, MAMO is 5.279, PSO-GWO is 5.854, and PL-ALO is 3.099.

461

Authors: Poornima Krishnamoorthy Balasamy, Kangaiammal Athiyappagounder

An effective E-Learning system must identify learning content appropriate for the needs of the specific learner from among the many sources of learning content available. The learning content can consist of different types of media content namely Text, Video and Image and to determine a rating. Recommendation systems become increasingly used in information systems and especially in e-learning platform. These systems are used to propose and recommend content of these platforms to users according to needs of the later in order to allow them to have the maximum information for learning. In this paper, a recommendation system based on data mining and deep learning has been proposed to help the learner by recommending the appropriate learning content. To improve the recommendations a content matrix is created and based on this, the logistic regression and deep learning methods could classify recommendations. Experimental results show that the precision, recall and f measure for Top 5 & 10 Recommendations using Deep Neural Network performsby 1.439% better than Top 5 & 10 Recommendations using Logistic Regression among Chi Squared, SA, PSO and ICA methods.

483

Authors: Amandeep Kumar, Balwinder Singh Dhaliwal, Damanpreet Singh

In recent years, researchers have focused their efforts on improving agricultural production by automating farming applications using wireless sensor networks (WSNs). The agricultural information must be accurate and updated in order to take the necessary actions. Sensor nodes (SNs) deployed in agricultural fields are power-constrained and require lot of energy to transmit data on regular basis. As a result, robust approaches are required to manage energy-related concerns for long-distance communication in large farms. To address long-distance agricultural needs, a cross-layer-based energy-efficient wireless sensor network model called CL-EEWSN that employs the wild horse optimizer (WHO) has been proposed. The primary goal of this research is to select the optimal cross-layer based cluster heads (CH) in terms of energy efficiency, communication delay, and latency. The physical layer, medium access control layer and network layer properties of each SN are used to evaluate and select the best CH. Further, a deep learning-based approach is used to determine the best data transmission path. The CL-EEWSN reduces energy consumption, end-to-end delay, and improves packet delivery ratio when compared to existing state-of-the-art protocols. The average energy consumption of CL-EEWSN for 100 nodes and 400 nodes is 0.16 nJ and 0.31 nJ respectively and is lowest in comparison to existing techniques.

505

Authors: Pradeep Krishnadoss, Chirag Chandrashekar, Vijayakumar Kedalu Poornachary

Cloud computing, a term which has become very popular on the internet platform which has evolved over time and gained more attention due to various features that are being introduced like multiplex numerous users on the identical physical structure, computing of resource on request and many more. Cloud computing is considered to be an enormous pool having essential resources which the users can make use of to complete their task through the internet. Therefore, in order to complete the tasks requested by the users, a scheduling algorithm is required making scheduling a vital part of cloud computing. Though numerous solutions with respect to scheduling techniques have been proposed in in recent literature, researcher’s continuously carryout performance upgradation by considering a mixture of QoS parameters thereby enhancing the cloud performance. In this regard, we had introduced an optimal and ideal task scheduling algorithm and challenge the other prevailing algorithms in terms of QoS parameters like makespan and energy. The proposed algorithm increases the performance of the scheduling in cloud by making use of an enhanced meta-heuristic hybrid algorithm called Rider Cuckoo Optimization Algorithm (RCOA), which is an advanced version of the already existing Rider Optimization Algorithm. The proposed RCOA technique had produced an improvement of 5.6%, 4.36% and 2.27% for makespan and had reduced the energy to the tune of 16.8%, 19.18%, and 16.15% when compared with Particle Swarm Optimization (PSO), Cuckoo Search (CS), Rider Optimization Algorithm (ROA) algorithms respectively when used with 25 VMs. Also the proposed approach attains the objective of optimal scheduling swiftly and achieves convergence in a small duration of time.

527

Authors: Emtinan Mohammed Jawad, Haidar J. Mohamad, Hazim G. Daway

Colour retinal image enhancement plays an important role in eye diseases and classification techniques. Optical coherence tomography devices are used to take images, which often suffer from a lack of lighting and contrast. In this study, a new algorithm for improving colour retinal images was suggested. The suggested algorithm was developed on the basis of improving the lighting (L) based on (Lab) colour space while preserving the colour information (a and b). The lighting enhancement included the wavelet fusion between the original lighting component and the improved component twice using contrast-limited adaptive histogram equalisation (CLAHE). Data containing 40 image forms (driver rental images dataset) were used, and the proposed method of improvement was compared with several other methods, such as principal component analysis using reflection model (PCAURM), luminosity and contrast adjustment (LCA), fuzzy logic by stretch membership function (FLSMF), CLAHE, contrast enhancement approach (CEA) and modified color histogram equalization (MCHE). The collected results showed that the proposed method obtained excellent quality rates in terms of natural image quality evaluator (4.680), entropy (6.985), contrast enhancement measurement (0.569) and structural similarity index (0.963).

548

Authors: Jaafar Ahmed Abdulsaheb, Dheyaa Jasim Kadhim

This paper discusses the best path planning algorithm for an autonomous mobile robot in an unknown environment with irregular static and dynamic obstacles and a static and dynamic target based on the improved COOT (ICOOT) optimization algorithm. The ICOOT overcomes the drawbacks of unstable searches in the conventional COOT. The path planning problem is solved by finding the collision-free path between multi-objective shortest path and smoothness. The ICOOT tries to imitate the real world by adding the mobile robot's actual size and the kinematic model with specifications for mobile robots. In order to evaluate the proposed algorithm, thirteen benchmark test functions are used to make a comparison with 30, 100, and 500 dimensions. To test the efficiency of the proposed technique, results are compared with five swarm optimization algorithms. The standard deviation results show that the proposed algorithm gets the best results in 84% of the thirteen test functions for 30 dimensions and 92% for 100 and 500 dimensions. Also, in four complex maps (10×10) m, the mean results show that this method is very useful for robot paths from the start to the target, and the mean distance for ten runs is 13.3797 m for map 1, 13.5164 m for map 2, and 11.9312 m for map 3, and 16.3937 m for map 4. It showed how well it could move quickly and easily around both fixed and moving obstacles.

579

Authors: Mohammed Jabbar Mohammed Ameen, Saad S. Hreshee

This paper introduces a novel voice encryption wireless system, which is based on characteristics of massive Multiple Input Multiple Output (MIMO) wireless channel. By exploiting the Minimum Mean Square Error (MMSE) precoding technique, channel fading values are permutated and substituted using various chaotic generators. The voice samples are mixed with channel values and a chaotic sequence before being transmitted. The proposed model was applied on a 5G network that included Generalized Frequency Division Multiplexing (GFDM), Parallel Spatial Modulation (PSM), and massive MIMO. Two chaotic sequences are generated, the first one used for the substitution stage by QR decomposition for the Henon map and then combining the outcome with Bernoulli and logistic maps to Q and R matrices, respectively. New chaotic signals are added to the real and imaginary parts of the MMSE precoding matrix. The second key is generated by the Tent map to perform the permutation stage for the overall precoding matrix. Several security metrics tests were used to verify effectiveness the of the proposed method resistance against attacks .The obtained tests results are: Signal to Noise Ratio (SNR=-23.1223), Root Mean Square (RMS=0.88565), Crest Factor (CF=10.6717), correlation coefficient (r_xy=0.000471), Linear Predictive Code Distance (dLPC=1.2758), Spectral Segment SNR (SSSNR=-27.6149), Cepstral Distance (dCD=9.4159), number of samples change rate (NSCR=33.334%), unified average changing intensity (UACI=99.9998%), keyspace=2500, key sensitivity, and speed=0.0329.The simulation results show that the proposed scheme has an excellent level of security and resistance to various forms of attacks, which outperforms many recent similar voice encryption methods.

601

Authors: M. Syauqi Hanif Ardani, Riyanarto Sarno, Mikhael Ming Khosasih, Malikhah Malikhah, Doni Putra Purbawa, Chastine Fatichah, Dwi Sunaryono, Rahadian Indarto Susilo, Shoffi Izza Sabilla, Kelly Rossa Sungkono

Human axillary odour is a unique discussion because it stores various useful information, including health, food quality analysis, and the similarity to the smell of the food eaten which can be done using an electronic nose (e-nose). However, rarely for research use an e-nose to determine the similarity of human axillary odour to food material odours. This research aimed to determine the similarity between male and female odours based on food material odours, such as garlic, shallot, and red chilli using an e-nose. We propose a method using an electronic nose with 8 Taguchi gas sensor (TGS) and 1 digital & humidity (DHT) sensor, then processing the resulting data using the fast fourier transform (FFT) smoothing method. Sensor data that have anomalies are removed by the quantile method. The standardization process is carried out to the signal data and feature extraction is processed with mean, standard deviation, and minimum value. This study found that the proposed method can produce a fairly good cluster of aroma food materials, faithfully getting the highest accuracy of 96.67 % using support vector classifier (SVC) and k-nearest neighbour (KNN) with each best parameter. We also found that the smell of human armpit sweat generally has a lot of similarity with food materials, specifically that male has the highest similarity to the shallot, while female has the highest level of similarity to garlic. In contrast, red chilli has the lowest level of similarity with human armpit odour.

625

Authors: Aldinata Rizky Revanda, Chastine Fatichah, Nanik Suciati

Manual classification of acute lymphoblastic leukemia carried out by doctors will certainly take a lot of time and effort. The challenges in automated computer-based systems for classification of acute lymphoblastic leukemia are when providing proper lightning in stained white blood cell microscopy images and when segmenting the touching or overlapping cells in the image. Previous studies related to the classification of acute lymphoblastic leukemia still require many steps when using conventional methods, whereas when using deep learning methods are still limited to classification without providing analysis for the instance segmentation of lymphoblast in the image. Therefore, we propose instance segmentation using Mask R-CNN on white blood cell microscopy images to classify acute lymphoblastic leukemia that can support the diagnosis process efficiently and effectively. In this study, we implemented Mask R-CNN by transfer learning method to fit the instance segmentation task on white blood cell microscopy images. We added a contrast enhancement process to the image dataset to overcome the bad lightning problem in stained white blood cell microscopy images. We used the real dataset obtained from hospital to evaluate our method. The method we used was able to get 83.72 % accuracy, 85.17 % precision, and 81.61 % sensitivity.

652

Authors: Weerapong Chanbua, Unnat Pinsopon

Steady state and dynamic LuGre friction models of an electro-hydraulic system were constructed. Indirect electro-hydraulic cylinder force control under friction compensated PID and fuzzy controllers were tested and compared. Feedback force signal was indirectly measured and calculated from pressures to emulate the applications that direct force measurement is impossible. Estimated friction was compensated to desired force command of the PID controller. The proposed fuzzy controller contained two modules. The first, as a friction compensator, calculated modified error according to estimated friction and feedback force error. The second module calculated control action based on PD control scheme. Friction compensated PID and fuzzy controllers could reduce tracking errors at the maximum force commands by 80% and 90%, respectively, compared to the PID controller without compensation. The steady state friction compensation yielded better tracking performances compared to the dynamic friction. Tracking performance of the fuzzy controller was always better than the PID controller.

676

Authors: Anik Vega Vitianingsih, Zahriah Othman, Safiza Suhana Kamal Baharin, Aji Suraji, Anastasia Lidya Maukar

The difficulty of acquiring data from numerous intergovernmental agencies/institutions for prone road traffic accidents (PRTA) spatial datasets produces a small-scale dataset that causes dataset imbalance. Class imbalance in small-scale datasets causes uncertainty in the results of the modeling PRTA classification. The proposed research is a scenario-based case representation model on the pre-processing data stage to increase the sensitivity of algorithm classification in a small-scale dataset that causes dataset imbalance using machine learning (ML), the synthetic over-sampling method. The retrieval of attributes from the spatial dataset is transformed into the raw dataset, the normalized dataset, the synthetic minority over-sampling technique (SMOTE) raw dataset, and SMOTE normalized dataset scenarios. Balancing datasets using four variants of SMOTE, namely ADASYN, Borderline-SMOTE, K-Means SMOTE, and SVM-SMOTE resampled. To evaluate how well the PRTA classification model performed, we utilized the hyper-parameters optimization technique and the genetic algorithm (GA) search cross-validation. The experiment was run with the ML classifier method, including the k-nearest neighbor (KNN), support vector machines (SVM), multilayer perceptron (MLP), naive bayes (NB), logistic regression (LR), and random forest (RF). The Area Under Curve (AUC) was used to evaluate the results of the experiments. The results of the dataset test in a predetermined scenario conclude that a single algorithm that is computationally light to produce an optimal classifier tends to use a raw dataset that is balanced using SMOTE. The KNN method as a single algorithm for classification based on the distance between samples is superior, with an AUC value of 0.89, which is included in the good classification category of all ML classifiers proposed to handle small data sets imbalanced classes using SMOTE raw datasets for K-Means variants SMOTE.