International Journal of Intelligent Engineering and Systems

1

Authors: Surya Prasada Rao Borra, Rajesh Kumar Panakala, Pullakura Rajesh Kumar

Nowadays, all the medical diagnosis is achieved by using the Digital Image Processing (DIP). Because, the usage of DIP is more important in the medical field to identify the activities of the patients related to various diseases. Magnetic Resonance Imaging (MRI) and Computer Tomography (CT) scan images are used to perform the fusion process. In brain medical image, MRI scan is used to show the brain structural information without functional data. But, CT scan image is included the functional data with brain activity. To improve the low dose CT scan, Hybrid algorithm is introduced in this paper which is implemented in FPGA. The main objective of this work is to optimize performances of the hardware. This work is implemented in FPGA. The combination of Discrete Wavelet Transform (DWT) and Principle Component Analysis (PCA) is known as hybrid algorithm. The Maximum Selection Rule (MSR) is used to select the high frequency component from DWT. These three algorithms have RTL architecture which is implemented by Verilog code. Application Specified Integrated Chips (ASIC) and Field Programmable Gate Array (FPGA) performances analysed for the different methods. In 180nm technology, DWTPCA- IF architecture achieved 5.145mm2 area, 298.25mW power, and 124ms delay. From the fused medical image, Mean, Standard Deviation (SD), Entropy, and Mutual information (MI) performances are evaluated for DWT-PCA method which has better performance than conventional methods.

22

Authors: Pundru Chandra Shaker Reddy, Alladi Sureshbabu

India is a country which has exemplary climate circumstances comprising of different seasons and topographical conditions like high temperatures, cold atmosphere, and drought, heavy rainfall seasonal wise. These utmost varieties in climate make us exact weather prediction is a challenging task. Majority people of the country depend on agriculture. Farmers require climate information to decide the planting. Weather prediction turns into orientation in the farming sector to decide the start of the planting season and furthermore quality and amount of their harvesting. One of the variables are influencing agriculture is rainfall. The main goal of this project is early and proper rainfall forecasting, that helpful to people who live in regions which are inclined natural calamities such as floods and it helps agriculturists for decision making in their crop and water management using big data analytics which produces high in terms of profit and production for farmers. In this project, we proposed an advanced automated framework called Enhanced Multiple Linear Regression Model (EMLRM) with MapReduce algorithm and the Hadoop file system. In the proposed model (EMLRM) first, we stored the unstructured weather data in hadoop distributed file system (HDFS), process that stored data by using MapReduce Algorithm and build the rainfall prediction model by utilizing Multiple Linear Regression. We used climate data from IMD (Indian Metrological Department, Hyderabad) in 1901 to 2002 period. The experimental outcomes show that the EMLRM provided the lowest value of Root Mean Square Error (RMSE= 0.274) and Mean Absolute Error (MAE= 0.0745) compared with existing methods. The results of the analysis will help the farmers to adopt effective modeling approach for predicting long-term seasonal rainfall.

43

Authors: Aditi Jahagirdar, Manoj Nagmode

Extracting keyframes for action recognition purposes is a challenging task, as compression is to be achieved without losing the impression of the action. Insufficient or wrong keyframes can lead to confusion in the type of action. Proposed two-level key frame extraction algorithm uses an adaptive threshold technique to identify most dissimilar frames as keyframes. At the first level, global features based on intensity histogram and at second level local features computed from wavelet decomposition are used as similarity measures. A new performance parameter, Compression Ratio-Normalized Fidelity (CRNF) is introduced for evaluating the keyframe extraction algorithm. Average CRNF of 0.81 is achieved by the proposed algorithm, as compared 0.37 achieved by existing histogram-based method for action recognition dataset. CRNF value of 0.95 is achieved for Open Video project dataset, which is more than existing methods. Qualitative results further prove the effectiveness of the proposed algorithm.

63

Authors: Aditya Kumar Sahu, Gandharba Swain

Since the inception of the reversible data hiding (RDH) concept, it has been a compelling topic in the field of data hiding. Being reversible, it has the ability to restore the original image followed by the successful retrieval of the secret data, at the receiving side. The concept of the dual stego-image based RDH technique utilizes two identical images of the original image for concealing the secret data, has gained wide compliance. Therefore, someone with both the stego-images can only extract the concealed data. In this paper, two improved dual imaging based RDH techniques, such as (1) dual stego-image based pixel pair LSB matching with reversibility, and (2) dual stego-image based modified LSB matching with reversibility, are proposed. In technique 1, at first two mirrored images are obtained from the original image. Then, using the pair of two consecutive pixels from the original image, the mirrored images pixels are modified using LSB matching technique. Later, these pixel pairs are readjusted to ensure reversibility at the receiving side. Similarly, technique 2 utilizes each original pixel to generate two distinct stegopixels using modified LSB matching. The experimental result shows that the technique 1 maintains excellent peak signal-to-noise ratio (PSNR) of 51.29 dB and 51.30 dB for the two stego-images with hiding capacity (HC) of 524288 bits. At the same time, technique 2 offers 51.19 dB and 49.44 dB of PSNR while exhibiting the equal HC. Further, investigation with various image quality assessment (IQA) metrics like quality index (QI), and structural similarity index (SSIIM) are proven to be competent over the other existing works considered in this paper. In addition, both the proposed techniques have shown excellent anti-steganalytic ability against RS and pixel difference histogram (PDH) attack.

85

Authors: Mahmood Hamza Al-Mufraji, Thamir R. Saeed, Ansam Subhi Jabbar, Jafar W. Abdul Sadah

In many applications, in industry and laboratory, there is a robot with a human, in close motion, to satisfy the work requirement. Therefore, for safety and monitoring the quality, the classification between them is important. The robot is considered as a semi-rigid object because it has many parts in the half upper, while the lower half is rigid, and a human is a non-rigid object. This paper presents a practical result which investigates the classification between these two objects based on the micro-Doppler signatures. The object's data were acquired using an S-band 2.4 GHz radar. The improved Stockwell transform was used to analyze the radar received signal in the timefrequency domain and feature extraction. The singular value decomposition is used for data filtering based on the Hankel matrix (HSVD) while using the Temporal Radial Basis Function to satisfy the classification. The classification accuracies gained are; 96.4%-100% for semi-rigid, 92%-95% for non-rigid closed objects.

104

Authors: Tejas Mehta, Chandu Bhensdadia

In this paper, we present an adaptive approach in order to match and retrieve near duplicate images at different scales. Matching only local Features does not necessarily identify visually similar images. Global features are fast at matching but less accurate. Many existing methods either use local features or CNN features for image or video retrieval task. In this paper, we combined the use of SURF local points and CNN features extracted around SURF points in order to match near duplicate image pairs. Image pairs are segmented into blocks and CNN features of the image block containing matched SURF features are extracted and matched. Regions around matched image blocks are grown adaptively and matching is carried out until CNN mismatch is observed. To verify our proposed approach, experiments are carried out on benchmarking California-ND and Holiday dataset. Compared to traditional approaches for image retrieval, our approach not only retrieves relevant images but also provides detail of localized matched patch. For California-ND dataset and Holiday dataset, we achieve remarkable mAP (mean average precision) score up to 0.86 and 0.74 respectively.

130

Authors: Nagendar Yerukala, Venu Nalla, Padmavathi Guddeti, V. Kamakshi Prasad

Data encryption play major role in all communications via internet, wireless and wired media. Stream ciphers are used for data encryption due to their low error propagation. This paper presents a new design of stream cipher for generating pseudorandom keystream with two LFSR`s, one FCSR and a non-linear combiner function, which is a bit oriented based on alternating step generator (ASGF). In our design two LFSR`s are controlled by the FCSR . ASGF has two stages one is initialization and the other is key stream generation. We performed NIST test suite for checking randomness on the keystream of length 1500000 generated by our design with 99% confidence level. Keystream of ASGF passes almost all NIST tests for randomness and the results are tabulated. For fixed extreme patterns of key and IV, ASGF is giving random sequence. Throughput of our proposed stream cipher ASGF is 4364 cycles/byte. Throughput comparison of ASGF with existing stream ciphers A5/1, A5/2 and RC4 are presented. It is not possible to mount brute force attack on ASGF due to large key space 2^192. Security analysis of ASGF against exhaustive search, Algebraic attack, distinguishing attack is also presented in this paper.

149

Authors: Ali Mohamed Riyad, Mohammed Saleem Irfan Ahmed, Husni Hamad Almistarihi

An intrusion detection system is one of the network security tools installed to monitor suspicious activity in the network and act as a last line of defense. It normally notifies about the skeptical activity occurred in the network using sensors by sending alarms to the administrator. However, the IDS present in the large network generates not only a large number of alerts but also abundant false alerts. These generated alerts are very difficult to handle as it increases the burden for the network administrator and also pulls down the performance of the defense system. In order to overcome the issue, various countermeasures have been proposed. Commonly, to increase the quality of alerts, the alerts are post-processed in such a way that the false alerts are filtered out thereby refining the performance of the IDS defense. In this paper, we propose an IDS quality framework using alert post-processing techniques to separate out the false alerts generated by various sensors in the network. At low level alert post-processing, the priority scores are assigned based on the quality measures to filter the irrelevant alerts having less significance. At high level alert postprocessing, higher level operations such as alert aggregation, clustering, and hyper alert correlation have been carried out to minimize the number of alerts and the high level report consisting of significant alerts is presented to the administrator. Experiments have been conducted using DARPA 2000 dataset to assess the performance of the proposed system. The system has produced pleasing results than many of the existing methods with 95% of alert reduction rate, 99% of completeness and 100% of soundness towards enlightening the quality of the alerts generated by the IDS.

170

Authors: Amsaveni Muthusamy, Duraisamy Subramani

Attribute reduction for big data is an important preprocessing step in the area of data mining. A multi-step dimension reduction approach was proposed for attribute reduction in big data. It addressed the non-linear relationships within the attributes. The data dimension was reduced through a parametric mapping. The mapping parameters were estimated using low-rank Singular Value Decomposition (SVD). However, the user-defined criterion in multi-step dimension reduction approach has greatly influenced the efficiency of attribute reduction. This approach was proposed for a single machine that means the entire big data must fit in the main memory and the parallelism was limited. So, in this paper, parallel rough set theory based attribute reduction approach is proposed for attribute reduction in big data. Based on two descriptions of lower approximation and upper approximation, a rough set is constructed. Then a reduct is detected using inner importance measure and outer importance measure. The rough set theory is used in MapReduce framework to achieve the parallelism for attribute reduction in big data. Hence, the computation time is reduced by using parallel rough set theory based attribute reduction approach. Finally, the experiments are carried out in Amazon customer review, REUTERS-21578 and International Cancer Genome Consortium (ICGC) on AWS datasets to prove the effectiveness of parallel rough set theory based attribute reduction in terms of accuracy, precision, recall and computation time.

190

Authors: Ari Firmanto, Riyanarto Sarno

Product reviews can be used as suggestion for companies to improve their services in this digitalization era. These reviews can be presented in more detail using aspect based sentiment analysis. In this research, aspect-based sentiment analysis method is proposed using grammatical rules, word similarity, and SentiCircle. This method began with extracting the candidate aspects rules based on phrase detection in constituency parse. The candidate aspects were categorized using word similarity. Word similarity calculated the similarity value between the candidate aspects and the keywords obtained from Wikipedia. To determine sentiment polarity, SentiCircle is used. SentiCircle can capture the contextual sentiment from the data. The result showed that the proposed method was able to categorize aspects correctly, with the highest f1-measure value of 84%, while sentiment analysis produced the highest f1-measure of 87%.

214

Authors: Vemula Vinay Kumar, Kodali Sai Krishna, Sarukonda Kusumavathi

Brain tumour segmentation is a significant task in the Medical Image Processing (MIP). Early diagnosis of the brain tumours plays an important role in improving treatment possibilities and maximizes the survival rate of the patients. Manual segmentation of the brain tumours for the cancer diagnosis form large amount of MRI images that generated in clinical routine, its complex and time consuming task. So, the automatic brain tumour image segmentation process is required. In this research, a tumour portion is segmented from the brain image by using Adaptively Regularized Kernel-Based Fuzzy C-Means (ARKFCM) algorithm. The input images are resized like 256×256 in the pre - processing stage. The pre-processed MRI image segmented by ARKFCM, which is a flexible high-level Machine Learning (ML) technique to locate the object in a complex template. Next, Hybrid Feature Extraction (HFE) performed on the segmented image to improve the feature subsets. The feature selection process was performed by Kernel Nearest Neighbour (KNN) based Genetic Algorithm (GA) in order to acquire the best feature values. The best feature values given to the Deep Neural Network (DNN) classifier as an input, which is classified into Meningioma, Glioma and Pituitary regions in the MRI images. The performance of proposed ARKFCM - HFE - DNN method is validated by T1-WCEMRI dataset. The experimental outcome showed that the ARKFCM - HFE - DNN method improved the average classification accuracy up to 6.19 % than existing classification techniques such as Convolution Neural Network (CNN) and Support Vector Machine (SVM).

232

Authors: Bellam Purushotham, Choda Durga Venkata Subba Rao

Ad hoc networks are the infrastructure-less networks that are easily deployable and self-configuring, with each constituent node performing like a transceiver. The nodes are free to move subjectively in Mobile ad-hoc networks (MANET). Thus, the network’s wireless topology may be haphazard and may alter rapidly. The main important aspect to be handled is choosing a proper path for transmitting data in case of Link Failure (LF). The main concern of the research work is maintaining the path stability in the occurrence of LF in the network. Ant Colony Optimization (ACO) is used for optimizing the Routing Protocols (RP) such as Ad-hoc on demand Distance Vector (AODV). The Prediction of LF (PLF) mechanism is used for choosing the stable/proper path in the occurrence of node LF in the maintenance phase of distance vector RP based on Neighbour Node (NN). Hence, AODV-ACO-PLF is used for efficient transmission of data packets in the ad-hoc networks by maintaining stable path in-case of LF. The path stability is also maintained in presence of LF in the networks by choosing a different path using PLF Prediction Mechanism. The “AODV-ACO-PLF” methodology is used for finding the ideal way to communicate with the source and destination by giving better results in End-to end delay (6%), drop (5%), packet delivery ratio (7%), and energy consumption (7.5%) than existing methodology such as Path Selection Algorithm based on Hamming Distance and delay parameters (PSA-HD) and Predict Failure in AODV (PF-AODV) methodology in MANET.

249

Authors: Panya Jirajaracheep, Tanairat Mata, Pisit Boonsrimuang

In this paper, the overlapping Partial Transmit Sequence (overlapping-PTS) with Artificial Bee Colony (ABC) phase optimization method is proposed to reduce the larger Peak-to-Average Power Ratio (PAPR) which is the major drawback of filter bank multicarrier with offset quadrature amplitude modulation (FBMC-OQAM) system. In this proposed method, we clarify the problem of larger PAPR by reducing the PAPR of each overlapping data symbol with lower combination of the optimal phase factors. Unlike the conventional PTS methods, the proposed method uses the lower combination of phase factors by applying the ABC algorithm with considering the overlap between the current data symbol and the past data symbol to search the optimal phase combination. The proposed overlapping-PTS with ABC phase optimization method can significantly reduce the larger PAPR with low computational complexity for the FBMC-OQAM system which can be verified by the computer simulation results. From the computer simulation results, it can be confirmed that the proposed method shows much better PAPR which can reduce the PAPR by approximately 1.9 dB at CCDF 10-1 as comparing with the conventional PTS method and can perform much lower computational complexity than the conventional methods which are only 4.38% and 0.76% for S phase factor pattern =300 and 50 respectively.

267

Authors: Muthuselvi Gomathinayagam, Saravanan Balasubramanian

Demand Side Management (DSM) provides a better solution in order to manage increased electricity demand in the power system network. The DSM program relieves the stress on the electrical network for maintaining power system reliability during peak hours. This work proposes a new scheduling approach based on an Adaptive Differential Evolution algorithm (ADEA) by considering a new recombination probability factor (CP) and real mutation factor (F) for analysis. This proposed method is analyzed for industrial, commercial and residential network loads. The main aim of this demand control technique is to reduce the difference between the target curve and total load consumption curve. This paper provides a better solution compared to other algorithms like Evolutionary Approach (EA) and Symbiotic Organism Search (SOS) to reduce the peak load and electricity cost in industrial, commercial and residential distribution networks. The proposed method gives 8.2 % peak load reduction compared to EA for residential area and 2.28% reduction in peak load when compared to SOS for commercial sector. Also it reduces the electricity cost at 17.25%, 20.76% and 21.15 % for residential, commercial and industrial sectors when compared to without DSM. The participation factor of a consumer may be increased by price factor and the less violation in their scheduled demand. The described concept provides a relatively accurate fit to the target curve after load shifting. This concept will increase consumer participation in the DR Program.

287

Authors: Archana Nagelli, Ramesh Ragala, Badarudeen Saleena

Nowadays, the Extreme learning machine (ELM) is playing a key role in machine intelligence and big data analytics due to its various advantages such as fast training rate, universal classification/regression and the capability of approximation. The standard ELM uses the Moore–Penrose generalized pseudo-inverse for solving the hidden layer activation matrix and also it identifies the output weights. Because of that, the standard ELM takes more time to train the features from the dataset. In ELM, scalability also considered as a one of the major concern while processing the large dataset. In order to overcome this concern, the Automotive Rank based ELM (AR-ELM) is proposed to obtain an effective tensor decomposition for diminishing the training time. Besides, the Bayesian approach is considered in this AR-ELM to remove the redundancy from the decomposed samples of the tensor. The major objective of this proposed AR-ELM is to process the large amount of dataset without depending on memory capacities. The recognition accuracy is improved by eliminating redundant information. The key idea of the ARELM is to reduce the training time while processing the huge dataset. The implementation and simulation of the ARELM is done in Spark Python 3.7. The performance of the AR-ELM is analysed in terms of accuracy, precision, recall and training time. The proposed methodology is compared with three existing methodologies such as basic ELM, ELM-TUCKER and ELM-PARAFAC. The recognition accuracy of the AR-ELM methodology with Hardlims activation function is 0.8879 for letter recognition dataset, it is high when compared to the basic ELM, ELM-TUCKER and ELM-PARAFAC that are 0.8102, 0.8375 and 0.834 respectively.