Intelligent Data Communication Technologies and Internet of Things

Information Networks (INs) are abstract representations of real-world interactions among different entities. This paper focuses on a special type of Information Networks, namely Heterogeneous Information Networks (HINs). First, it presents a concise review of the recent work in this field. Then, it proposes a novel method for querying such networks, using a bi-functional machine learning algorithm for clustering and ranking. It performs and elaborates on supervised and unsupervised, proof-of-concept modeling experiments on multi-typed, interconnected data while retaining their semantic importance. The results show that this method yields promising results and can be extended and utilized, using larger, real-world datasets.

Adulteration and the introduction of counterfeit medicine in the pharmaceutical market is an increasing problem in our country. Official reports from WHO states that nearly 35% of the fake drugs in the world are being imported from India making up to a 4000 crore drug mafia market. The main reason is human intervention at every stage of the stringent process in the current system. This induces inevitable human error which is taken advantage of. The absence of an automated system is the main and the biggest drawback fof the currently implemented system. The paper proposes an automated system that provides a unique recognition mechanism by generating machine-readable code for each entity. This generation of the machine-readable code is backed up by encryption and hashing and hence cannot be replicated by a third party.

In the digital world, adequate data has been stored and accessed under a cloud by using various users. The system emanates beneath the user management to resolve the security and privacy risks. In multi-user setup, the attendance to add operations carried out to sign-up and user cancellation from cloud to yield the risk overload. Further to avoid this problem, the paper castoff a Generalized-signcrypt identity-based encryption (Gs-IBE) with a random modulator that has been experimented. The conditional variable Am(r) been castoff to recognize the collusion in the random value. The Am(r) perform the distinct job in the random bit generation of each random element selection.

In today’s world, most people are intensely dependent on online social networks (OSN). People use social sites to find and make friends, to associate with people who share comparable intrigue, trade news, organize the event, exploring passion in an. According to a Facebook review, 5% of monthly active users had fake accounts, and in the last six months, Facebook has deleted 3 billion accounts. According to the Washington Post, Twitter has suspended over 1 billion suspect accounts over a day in recent months, Detection of a fake profile is one of the critical issues these days as people hold fake accounts to slander image, spread fake news, promotes sarcasm that has attracted cyber criminals. There are numerous machine learning methodologies such as supervised learning, SVM-NN, are produced for the effective detection of a fake profile. In this paper, convolution neural networks is proposed with many artificial neural network algorithms like face recognition, prediction, classification and clustering for the efficient identification of account being real or fake and elimination of fake profile account. Furthermore, the study is grounded on the fact of the face-recognizing of the user and performing feature detection and time series prediction. If the user account is detected fake it would not be created.

Cardiovascular diseases are one of the prime reasons for individual deaths across the globe, claiming millions of lives every year. Cardiovascular disease diagnosis is a critical challenge in the health care field as it has a lot of risk factors associated with it. Moreover, neural networks are ideal in making significant clinical decisions from the huge health-care data produced by the hospitals. This work is an effective method to find significant features and use Deep Neural Networks to build a cardiovascular disease diagnosis system. The proposed system is developed by using a well known dataset called Cleveland dataset of the UCI Repository. The model was introduced with different combinations of features and a deep neural network. The performance of the different subsets of features were evaluated and it was also investigated for a single gender. The proposed system with a minimal feature set helps in the early diagnosis of heart disease and aids the cardiologist in making clinical decisions.

Hepatic or Liver disease cause illness because of the perturbation of liver function. The liver performs many critical functions and if it gets injured or diseased, cause significant damage or even death. Failure to take care of early-stage hepatic issues will further deteriorate the organ to more complicated scenarios. Therefore the need to diagnose the condition early so that the condition has maximum potential for successful treatment. AutoML is the next coming wave of machine learning. In the market there are various commercial and open-source tools are available based on AutoML.Auto-WEKA package is such a tool and it is based on AutoML technique, which could be utilized by both experts, and non-experts group of users. The main usage of AutoML is to help the developer by automating the model selection and hyper-parameter optimization. Non-experts utilize AutoML with minimal code or without writing a single line of code. Tools that bases on AutoML make machine learning pipeline building effortless. In this paper, the implementation of Auto-WEKA for hepatic disease detection have been used.

An IoT-based continuous emission tracking and warning system are proposed which consists of Arduino ATMEGA328 processor, a Wi-Fi ESP8266 interface, gas sensor MQ135, LCD module and BYLNK application. It is placed at the exhaust of the vehicle to continuously monitor the exhaust characteristic data and is collected in the cloud server for further processing and alerting the vehicle owner. The prototype is monitored for the emission characteristics of BS6 standards. The BS6 standard is taken as reference as the existing BS4 standards have been withdrawn since April 2020. The objective of the paper is to introduce a vehicular emission monitoring and warning system and to ensure the pollutants limits are within BS6 standards. The prototype is capable of detecting and monitoring the pollutants such as CO, CO2 and NOx which are majorly found in the exhaust fumes. The measured data is shared through LCD and a BLYNK notification to the vehicle owner to continuously monitor the exhaust. This system is real-time, Portable, low cost and it provides a good interface with the application resulting in controlling the emission especially in the urban areas.

Secured network is an important issue in today’s era. The various defense techniques have been developed to protect the network from the attacks which hampers the performance of the network. The attacks like Blackhole, Identity Spoofing, Sybil, etc. are observed. In such types of attacks, the Sybil is the destructive attack. In Sybil attack, attacker misleads the other nodes by viewing the wrong identity of the users who are alert from the nodes in the network. In a Sybil attack, a node illegitimately claims various identities. Sybil attack intimidates network in routing, fair resource allocation, data aggregation and misbehavior detection. This paper focuses on a systematic study of Sybil attack and gives the critical analysis of the various parameters which affects the performance along with the advantages and limitations.

Recently, face recognition’s popularity has grown beyond imagination and has given a new dimension to the existing security system. From being theoretically proposing the idea on paper to making it a reality, it has gone through a lot of phases. This paper implements ease of access security system to the disabled people. The security is provided in terms of sending a text message to the registered mobile number and buzzing the alarm. The system is trained with images of only known faces of people for whom the door will be unlocked automatically. When the unknown person tries to enter the access to the door is denied. With leading-edge innovations, smart door unlocking system has become more push on it should be well designed to as it is related to home security and impart easy access to the user. This autonomous door unlocking system concerns to keyless door unlock system with the help of face recognition and detection.

The design of a frequency reconfigurable antenna placing PIN diode at the ground plane is proposed here. The geometry consists of a circular patch is mounted on an FR4 substrate. The PIN diode is used to disturb the current flow by changing the length of the conductor thereby changing the area. The biasing process will be easier when the diode is placed at the ground plane. So the fabrication of the antenna becomes also easy. The proposed antenna is useful in WLAN 2.45 GHz and UWB applications.

The main objective of this paper is to acquaint the reader with a trailblazing pike into the reciprocity axioms of the multidimensional discrete Hartley transform (DHT) through the multidimensional discrete Fourier transform (DFT). Spin-offs for the reciprocity axioms of multidimensional DHT are tendered through the reciprocity axioms of the multidimensional DFT. This is due to the mathematical hookup betwixt the DFT and DFT. DHT is a real transform, whilst the DFT is a complex transform. Real transforms are posthaste in implementation in contrast to the complex transforms. Ergo, the axioms deduced for the reciprocity of multidimensional DHT in this paper, can be used in real time for the implementation of reciprocity axioms of multidimensional DFT at a brisk pace, by half of the cent percentage than their actual use. Thence, the axioms for multidimensional DHT can be used in real-time applications that use multidimensional DFT such as dıgıtal video processing, colour image processing.

Cancer is one of the deadly diseases worldwide. It takes a lot of time and effort to detect it at the early stages. Various pathological and imaging techniques are widely used by doctors/physicians to ensure the presence of cancerous cells in the body. The techniques currently being used are highly time-consuming and lack accuracy. This paper deals with applying deep learning algorithm to detect the presence of different types of cancer. Brain, lung, and skin cancer are determined using Convolution Neural Networks (CNN). By utilizing pre-trained weights of VGG16, the model is trained and fine-tuned with cancerous and normal CT scan images. Experimental evidence shows that the adopted transfer learning (such as using pre-trained weights of VGG16) method provides better accuracy than traditional methods. The accuracy obtained is greater than 95%.

Video cameras are the most popular way of security used by the majority of people in homes and offices. These systems require a human being for video surveillance so that if some unusual activity happens, appropriate action can be taken. Usually, cameras are installed to monitor the unusual activities which may be a threat to the workplace. In the case of home security, the major reason to install video cameras is to monitor that no unknown person should enter the home without permission. This demand for home security systems has been utilized to provide automated surveillance security system which can specifically be used in such places to monitor the presence of an unknown person in the premises so that the people using this system are alerted using notification. For identification of unknown face, our system has been trained on some known faces using LBPH face recognizer and then the threshold is applied to decide whether the face is known or unknown, finally an alarm is generated if the face is declared unknown. The above system works well on good quality videos and images.

Deep Convolutional Neural Network learns with various levels of abstraction and has made a revolution among learning algorithms. Deep Learning currently plays a vital role in object classification, natural language processing, genetics, and drug discovery. The deep learning unravels the patterns by computing gradients to minimize the loss function and based on it the internal parameters are tuned to compute the layer-wise representation. The Deep Convolutional Neural Network has revolutionized computer vision and image processing. The work cross dissects the deep convolutional neural network to light its learning mechanism and extensively experiments parameter tuning to facilitate intelligent learning. The work comes up with a roadmap to build and train a deep convolutional neural network after extensive experimentation using CIFAR 10 dataset. The work also comes up with near-optimal hyperparameters that effectively generalize the learning of Neural Network. The performance of the Deep Neural Network is also evaluated with hypertension dataset gathered from the health department. On experimentation, could be inferred that the proposed approach gives comparatively higher precision. The accuracy of the proposed approach is found to be 90.06%.

The methods and approaches to language teaching and learning have always been changing with time from the Grammar Translation Method to the Communicative Language Teaching in order to be compatible with the globalized world. The governments across the world have realized the importance of implementation of ICT in education and especially in the Foreign Language Teaching and Learning. Researchers and policy makers also emphasize on the importance of integration of Information and Communication Technology in English Language Teaching. The status of ICT in education in Bangladesh is at the emerging stage where the adoption differs from the implementation and hence, it leads to a significant gap between the adoption and the actual implementation and outcome of ICT in education. Therefore, the present study aims to discuss the implementation of CLT in Bangladesh and to explore the impact of ICT in CLT. Further, it also attempts to highlight the challenges in the integration of CLT and in the incorporation of ICT in CLT.

It has been established and accepted that mental disorders pose one of the most prominent health challenges worldwide. Information retrieval from mental health data which may be explicit in electronic health records and clinical notes or may be implicit in social media postings has vast potential to detect, distinguish and diagnose the status of mental health of individuals and aid in managing this problem. This paper summarizes some recent studies that apply the state of art Artificial Intelligence (AI) techniques to mental health data. The paper summarizes that newly emerging AI technologies hold a decent promise and can be leveraged to predict, assist in the diagnosis and management of mental disorders. The role of AI in this area becomes particularly important in a scenario where there is a worldwide dearth of qualified professionals who can deal with mental health disorders, where the cost of these services is high and the people suffering from these problems often refrain from availing these services due to social stigma associated with it.

Frequent itemset mining (FIM) is among widely known and essential data analytics techniques, to discover and extract frequently co-occurring items. However, due to the massive information available online, it is difficult to extract valuable data with the help of FIM algorithms. Traditional FIM suffers from scalability, memory and computation issues. Onto this context, MapReduce framework can be used which can handle those issues along with algorithmic parallelization. In this paper, algorithms have been explored and compared majorly on two approaches for the reduction of the computation cost. The first approach for mining frequent itemset is candidate generation and data pruning; it is adopted by Apriori, equivalence class transformation and Sequence-Growth algorithm. The second approach for mining is through pattern growth, and the algorithm under this is FP growth. The parameters taken into comparative analysis are search strategy and load balancing.

In the current online advertising scenario, lookalike methods of audience expansion are highly valued to enhance the existing advertising campaigns. Machine learning approaches are highly favored, and the typically used models involve distance-based, keyword-based and classification techniques. In this perspective, the proposed research work has incorporated a comparative study of the basic distance-based clustering model for lookalike classification against a positive unlabeled classification model. The results demonstrate a greater accuracy of the positive unlabeled classification method.

In recent years, the growth of the Internet has led to the emergence of recommender systems, as embedded engines, in various domains. These engines are in a way helping tools for users, as they suggest items and services according to users’ tastes and preferences. But these recommendations depend upon past user/item liking history. Non-availability of past user/item history leads to the situation of cold start. The cold-start problem is one of the major factors affecting the efficiency of recommender systems. This survey explores the cold-start problem from a domain perspective. The scenarios of cold start in different domains and techniques used to find solutions have been explored. Most of the researches have proposed solutions for e-commerce/e-business sites as they have more economic value. In recent years, deep learning techniques have gained momentum and are being used for solving cold start. Solutions for this problem in domains other than e-commerce/e-business need to be explored using deep learning methods.

This paper discusses the application of image classification that can be used in controlling a car, unmanned automatically, to pause, move, take deviation as and when any obstacles come on its way, also follow traffic signals. Our work will focus on making cars to avoid accidents, unlike the manual ones that are prone to accidents. The reasons may be many lapses on the part of the driver physically, mentally or sudden medical emergencies, etc. As part of this work, CNN classifiers have been used to perform image classification. Using CNN there was accuracy of 90%. Further, this accuracy may be increased to over 95%. This work involves the use of OpenCV that controls the hardware such as Raspberry Pi, four DC motors, Raspberry Pi cam, Arduino Uno.

In the last decade, emergency responders, government organizations and disaster response agencies have certainly acknowledged that microblogging platforms such as Twitter and Instagram can be an important source of actionable information at the time of disaster. However, researchers feel that analyzing social media data for disasters using supervised machine learning algorithms is still a challenge. During the first few hours of any disaster when labeled data is not available, the learning process gets much delayed. The first objective of this study is to explore the domain adaptation techniques by making use of labeled data of some previous disaster along with the abundance of unlabeled data that is available for the current disaster. The second objective is to apply these domain adaptation techniques on disaster-related imagery data from the microblogging platforms since imagery data has largely been unexplored as compared to textual content. To achieve these objectives, domain adaptation methods would be applied to classify the images of an ongoing disaster as informative versus non-informative. This study, for the first time, proposes a semi-supervised domain adaptation technique where the classifier is trained on three types of data, labeled data of the previous disaster, unlabeled data of current disaster and a small batch of labeled data of current disaster. Our experiments have been performed on Twitter images corresponding to three disasters. The experimental results show that there is an improvement in the accuracy of the classification model if a small batch of labeled target images is also added along with the unlabeled target images and labeled source images at the time of training. The experiment aims to make the best use of the labeled data of a previous disaster to analyze the current disaster without any delay for better response and recovery.

Transport plays a crucial and vital role in the living standard of a nation. The transportation regulatory authorities across the world are continuously facing issues in traffic congestion, severe increase in pollution level and rise in the impurity level of air quality index. To overcome these problems, there is always a demand of intelligent strategies in transport management through automated software and IoT-enabled mechanisms. Due to cost-effective ration and lack of awareness, these technologies are not fully adopted in transportation industries. The primary aim of this research is to highlight prior researches in transport management and identify the issues. Further, the issues are reduced through the proposed model which improves the performance and reduces congestion so that environmental impact and productivity could be increased.

Quantum of data is increasing in an exponential order. Clustering is a major task in many text mining applications. Organizing text documents automatically, extracting topics from documents, retrieval of information and information filtering are considered as the applications of clustering. This task reveals identical patterns from a collection of documents. Understanding of the documents, representation of them and categorization of documents require various techniques. Text clustering process requires both natural language processing and machine learning techniques. An unsupervised spatial pattern identification approach is proposed for text data. A new algorithm for finding coherent patterns from a huge collection of text data is proposed, which is based on the shared nearest neighbour. The implementation followed by validation confirms that the proposed algorithm can cluster the text data for the identification of coherent patterns. The results are visualized using a graph. The results show the methodology works well for different text datasets.

With advancements in X-ray technology, there is an increase in the number of digital images used in the diagnosis of a patient. Whether it be a simple X-ray, MRI, CT scan or even a photo taken from a camera, the rise in the use of digital images has increased sharply. Though this has eased the entire process, it has brought the threat of cyber-attacks and breaches. The proposed method bridges this existing gap by incorporating suitable security mechanisms to preserve the privacy and confidentiality of medical diagnostic information of an individual. The approach utilizes neural networks to perform image steganography and combines it with a cryptographic algorithm (RSA) to secure medical images. The proposed method uses a two-level security providing a lower loss of 0.002188 on medical images improving upon existing image steganography techniques.

Fast Fourier transform (FFT) and inverse fast Fourier transform (IFFT) are being applied in various fields of digital signal processing (DSP) applications because of the advanced technology of VLSI. In recent communication system, orthogonal frequency division multiplexing (OFDM) is the most important FFT and IFFT application. FFT/IFFT processors in these communication systems consume high power, rendering the system inefficient. OFDM-based UWB systems IEEE 802.15.4a employ 64 point and IEEE 802.11a-based systems employ 128/64 point and, therefore, we designed 128/64 point FFT/IFFT processor. The radix-25 algorithm that we used reduces the number of non-trivial multiplications. By implementing clock gating technique and a low-power multiplier, we propose to design a low power, reconfigurable, variable-length FFT/IFFT processor. The simulation of the design was done using ModelSim and power consumption has been analyzed with 180 nm CMOS technology using Cadence Encounter tool. The power consumed by the processor without clock gating is 58.27 mW and with clock gating is 2.13 mW.

A robust digital image watermarking scheme in the DCT domain for color images using support vector regression is presented in this paper. This scheme attempts to achieve a fair balance between imperceptibility and robustness using an array of techniques including machine learning. The host image is transformed from RGB to YCbCr color space which gives a luma component used for embedding the watermark. This scheme utilizes Lagrangian Support Vector Regression (LSVR) for learning image features; entropy for selecting the image blocks in which watermark bits are embedded; discrete cosine transform to de-correlate the luminance component pixels. The low-frequency coefficients scanned in a zigzag manner are used for training and testing of LSVR model, and the output acquired from LSVR is used for embedding the watermark. The proposed method is applied to general image processing dataset images and found to yield fairly good results as shown by the comparison with existing methods.

In the field of science, the area of leaf in progressed pictures plays an exceedingly significant portion inside the field of science. The distinguishing proof of leaf may be an uncommonly basic key to maintain a strategic distance from an overpowering hardship of resign and the sum of rustic things. Our paper broadens the organization approach to arranging and testing images by local binary pattern descriptor on the image and store Histogram of Oriented Gradients is related to the expelled LBP features, then divided into two specific classes: organized images and oriented images, tolerating the support vector machine. The organized illustration on the databases implies rate of revelation generally 95.25% on our very own dataset. The time complexity to boot essentially diminished and the methodology is found to be working well underneath diverse lighting up assortment conditions. Every so often plants and leaves are used to prepare drugs for the resolution of numerous human diseases, and the selection of the accurate leaf and plant is compulsory in the search for precise medication. So the advantage of our project comes here that will help producers of such medicines to locate the specific leaf and plant for the precise preparation of medicines and thereby avoid errors.

Machine learning can learn a complex system using a large amount of data and has the potential in predicting critical medical emergencies like congestive heart failure (CHF). In this paper, a machine learning algorithm is proposed for predicting CHF using convolutional neural network (CNN). The system uses a VGG16 model, which is dependent on ImageNet and utilizes ECG signal. The CHF information is obtained from the Beth Israel Deaconess Medical Centre (BIDMC) CHF dataset and also the typical dataset from the FANTASIA database. Various models of CNN were prepared and tested on the dataset to obtain the information about CHF or normal for a patient until the required accuracy is achieved. The performance is assessed based on the accuracy of the code and precision. The ECG signal of a patient recorded for 24 hours is utilized in this research. The experimental result obtained by this research has given accuracy of 100% on applying the VGG16 Model to the dataset.

Twitter is a social media platform where users post their opinions on various events, products, services and celebrities. Automated analysis of these public posts is useful for tapping into public opinion and sentiment. Identifying negative public sentiment assumes importance when national security issues are at stake or when critical analysis of a product or policy is required. In this paper, a method is introduced that classifies tweets based on their negative content, without any prior training. Specifically, an unsupervised negative sentiment analysis is presented using an aspect-based approach. Phrase and keyword selection criteria are devised after identifying fourteen valid combinations of part-of-speech tags listed in a prioritized order, that are defined as phrase patterns. A sliding text window is passed through each sentence of the tweet to detect the longest valid phrase pattern. The keyword indicating the aspect information is detected using a dependency parser. SentiWordNet lexicon is used for scoring the terms in the detected keyword and phrase combination. The scores are summed up for each sentence of the tweet and transformed nonlinearly by a modified sigmoid function whose output is in the range [−2, 2]; this value comes out to be negative for negative tweets. The utility of our method is proved by superior results as compared to the state of the art on the benchmark SemEval 2013 twitter dataset.

Anxiety and stress are the problems of all times. Extensively, technology and art go hand in hand. Music therapy has been a sought-after technique for restoring one’s mental health problems including anxiety, stress and blood pressure. The body symptoms such as palpitations, age, sex, sweat, EEG signals, blood pressure and heartbeat rate help in easy investigation of stress. Manual analysis of these parameters can be cumbersome and time-consuming. Hence, a more contemporary approach to aid the health specialists in the treatment of the patients is by using efficient music along with computing technology of machine learning algorithms. This work is a new product used to help health specialist treat patients with mental disorder using Indian classical music and machine learning. In this work, a dataset of mental disorder parameters and music datasets are mapped together as inputs. Music datasets include Indian classical ragas suggested by renowned health specialists. Machine learning algorithms, namely logistic regression, SVM, K-nearest neighbor (KNN), random forest (RF classifier) and decision tree classifiers, are used to train the models, and the accuracy performance of each model is compared. From the various testing datasets, the efficiency seen by our experiments that support vector machine (SVM) is best suited for our work which has an accuracy of 87.23%. To make it simple for health specialist and caretaker to handle this with ease, a web application using Flask, a python framework has been designed. The application enables the user to listen to the audio files and keep a track of one’s stress status.

Reducing the contaminated water and obtaining fresh oxygen is difficult in these environmental conditions. Water is in the form of vapour and moisture mixture in the air. It is generally noted that 30% of water is almost wasted. The moisture in air is used in the proposed system. This system considers moisture air from atmosphere and converts it directly into drinking water. Condense process is used in the proposed system to convert moisture into water droplets. This device also generates oxygen (O2) from water through electrolysis process. This system consumes less power. The system sends the status of the system to the mobile application.

The stock markets are considered to be the most sensitive and volatile financial institutions. Investment-related decisions are made by looking at historical data and observing the patterns, and sometimes these turn out to be profitable and sometimes not. Investment involves making predictions based on various factors and later combining all of these to conclude. The market as a whole is vulnerable to news and leaks which in turn decide the sentiment of buyers and hence directly affect the price of a given stock. Nowadays, machine learning techniques are being used to forecast the trend of a given stock. This paper presents a relative analysis of the prediction of the stock price using algorithms long short-term memory (LSTM) and auto-regressive integrated moving average exogenous (ARIMAX), without and with sentiment analysis. It has been observed from results that both algorithms attain a considerable improvement in the forecasting when sentiment analysis is applied.

Image is considered as important data in the medical field because usage of medical images for diagnosing the disease keeps on increasing in the modern digital medical field. These images need to be encrypted at the time of transmitting over an insecure network for maintaining integrity and confidentiality. This work aims to propose a medical image cryptosystem with less computational time based on the fusional model of the modified one-dimensional chaotic map. First, the proposed fusional chaotic model generates the chaotic sequence and then the generated chaotic series is used for encrypting the intensity of medical images. Simulations and security assessment are evaluated by applying statistical and differential attacks. Robustness against exhaustive and noise attack results indicates the strength of the developed cryptosystem. Comparison analysis illustrates that the proposed cipher has efficient and enhanced security than state of the art. Hence, the developed cryptosystem has absolutely opted of medical image applications.

Digital learning and the virtual classroom had an enormous influence during this new era of modernization, which has brought a revolution on the interested student to acquire at their own comfort as well as the desired pace of learning. But the important success factor of traditional classroom pedagogy, i.e., real-time content delivery feedback, is missing. Eventually, engagement becomes crucial to strengthen and improve user interaction. It constrained us to resolve this gap issue before, because of the lack of publicly accessible datasets. But, now since DAiSEE is the first multi-mark video grouped dataset which comprises student’s recordings for perceiving the client’s full of feeling conditions of boredom, confusion, engagement and frustration in wild, that makes it a benchmarked dataset for solving such kind of problems. In this paper, the model to employ on our convolutional neural networks is a custom Xception Model, a depth-wise separable convolution pre-trained on the billion of images along with multi-task classification layers on it to recognize the affective states of user efficiently. The proposed Xception network slightly outperforms the frame-level classification benchmarked by DAiSEE’s model.

SDN separates the control plane from the data plane and thus provides dynamic and flexible management of the network. Using a single controller in the control plane will create a single point of failure. To overcome this problem, multiple controllers are used in the control plane. Multi-controller architecture has two kinds of mapping between switch and controller, single mapping, and multi mapping. The single mapping maps each switch to a single controller. When one controller is down, then another suitable controller in the slave role needs a role change, and this will increase the recovery time. While in the proposed multi mapping technology, each switch is mapped to two controllers. When one controller fails or overloads, then the other controller takes over the switch. Since the other controller is also in an equal role, recovery time is better. QoS parameters for both single mapping and multi mapping are compared using SWITCH topology and Bandcon topology from Internet topology zoo. For switch topology, multi mapping has reduced packet loss by 38%, reduced end-to-end delay by 21%, and increased throughput by 36% in comparison with single mapping. For Bandcon topology, multi mapping has reduced packet loss by 43%, reduced end-to-end delay by 36%, and increased throughput by 45% in comparison with the single mapping.

Search engines are exceedingly dependent on the query auto-completion. Query auto-completion is an ongoing activity that puts forwards a group of words for every click dynamically. Query suggestions help in formulating the query and improving the quality of the search. Graphs are data structures that are universal and extensively used in computer science and related fields. The graph machine learning approach is growing rapidly with applications such as friendship recommendation, social network, and information retrieval. Node2vec algorithm is used to study the feature illustration of nodes in a graph. It is derived by word embedding algorithm Word2vec. A supervised Recurrent Neural Network using Long short-term memory (LSTM) is employed to compute the accuracy. This model confirms 89% accuracy for query auto-completion. Greater the accuracy better the model.

This paper represents the relative differences between the previously proposed active inductor designs. The parameters such as quality factor, power consumptions, and inductance tuning have been considered while comparing the performances of these inductors, operating at radio frequency. Inductors are widely used in integrated circuits for their tunability property. RF applications of inductors include the use in filters, low-noise amplifiers, and VCOs. Over the years different designs of active inductors have been proposed to eliminate the problems associated with conventional on-chip inductors. Different designs that propose techniques to reduce distortions, to improve the quality factor and tunability have been considered in this paper.

In service computing, Quality of Service (QoS)-aware web service composition is considered as one of the influential traits. To embrace this, an optimal method for predicting QoS values of web service is implemented where credibility evaluation is computed by accumulating reputation and trustworthiness. An automatic approach for weight calculation is invoked to calculate the weight of QoS attributes; it improves WS QoS values. QoS value is optimized by using Genetic Algorithm. Services with high QoS values are taken as candidate services for service composition. Instead of just selecting services randomly for service composition, cuckoo-based algorithm is used to identify optimal web service combination. Cuckoo algorithm realizes promising combinations by replacing the best service in lieu of worst service and by calculating the fitness score of each composition. A comparative study proved that it can provide the best service to end-users, as cuckoo selects only service composition with high fitness score.

Emotions play a vital role in a human’s mental life. Speech is a medium through which expression of perspective and identification of one’s mental state is possible. Recognizing the feelings that others are trying to convey through speech is essential. There are various parameters of the speech signal that define the feelings of a person. Thus, speech emotion recognition (SER) from the speech signal is a challenging task. This paper proposed an SER system based on the features extracted and obtained by Mel-frequency cepstral coefficient (MFCC) spectrograms. As for a complex model, the audio features obtained by MFCC play an important role in emotion recognition. The 1D-CNN model architecture is implemented in this paper. The work is performed on “The Ryerson Audio-Visual Database of Emotional Speech and Song” (RAVDESS) dataset. Six emotions based on their gender are classified (i.e., 6 × 2 emotions) with 82.3% accuracy. The emotions classified are happy, sad, angry, calm, fear, and nervous.

The advancing COVID-19 pandemic caused by the novel coronavirus has taken the world by a storm due to its unprecedented nature. In order to increase the understanding of the disease and create countermeasures for the same, collecting and storing data in a proper and efficient format is of utmost importance. However, this tremendous amount of data is obtained from various heterogeneous sources and is usually dynamic in nature. Traditional RDBMS might not be the most efficient choice for the sporadic and ever-changing clinical data associated with COVID patients due to its highly rigid nature. This paper utilized a primary dataset acquired from COVID-19 patients as a premise to portray the inefficiencies of RDBMS and further proposes two new schemaless, unstructured databases, NoSQL and XML databases, as an offset to this drawback. The intention is to propose the two most efficient technologies and delineate the findings through a sample implementation.

“Frequent Itemset Mining” is a domain where several techniques have been proposed in recent years. The most common techniques are tree-based, list-based, or hybrid approaches. Although each of these approaches was proposed with the intent of mining frequent itemsets efficiently, as the number of transactions increases, the performance of most of these algorithms gradually declines either in terms of time or memory. In addition, the presence of redundant itemsets is another crucial problem where a limited investigation has been carried out in recent years. There is thus a pressing need to develop more efficient algorithms that will address each of these concerns. This paper aims to survey the different approaches highlighting the advantages and disadvantages of each of them so that in future effective algorithms may be designed for extracting frequent items while addressing each of these concerns effectively.

The research proposal discloses a novel microscopic device and method for making a colorless pathogen specimen (i.e., virus, bacteria) visible to the human eye. The novel device fits like a headgear on a user’s head. Further, the novel method involves the usage of evanescent waves, that are generated by surfaces of the objects that undergo excitation on the incidence of a certain light wavelength. The microscopic device uses optical sensors to detect the evanescent waves generated by the object surface. Further, the microscopic device also uses nano projecting units for projecting the captured evanescent waves on the projection screen that in turn magnifies the captured object’s evanescent waves. Hence, the disclosed research proposal serves as an effective solution for viewing tiny viruses that are not visible to the human eye, such as COVID-19 virus.

Social media keeps on increasing in size and demands automation in data analysis. Student shares their opinions, concerns, and emotions in the social Web site, because it has a variety of opinions that are central to most of the human activities and a key influence of behavior in their day-to-day life. Many of the tweets made by students have some sort of meaning, but some category does not have a clear meaning such as a long tail. In this paper, a different classification model is developed to analyze student’s comments which are available in social media. This paper mainly focused on emotions. Data is taken from 15,000 tweets of student’s college life and categories—study load of all majors, antisocial, depression, negative emotion, external factors, sleep problems, diversity problems. These multi-label emotional comments are to be classified, analyzed, and compared with the support vector machine and Naïve Bayes algorithm to show student learning problems. The experimental results show that major students’ learning problems make better decisions for future education and service to them.

Cloud computing is a recent buzzword in the field of information technology (IT) that has changed the way of computing from personal computing to virtual computing done with the help of cloud service providers via the Internet. It is evolved from distributed computing included with some other technologies such as virtualization, utility computing, service-oriented architecture, and data center automation. It is based upon the concept of computing delivered as a utility encapsulated with various other characteristics such as elasticity, scalability, on-demand access, and some other prominent features. To keep these properties intact during high demand for services, the load must be balanced among the available resources in the cloud environment. This load can be of various types such as CPU load, network load, memory load, etc., and balanced by executing a load balancing mechanism after detecting the overloaded and underloaded nodes. To achieve this researchers design different types of load-balancing algorithms for optimizing different performance parameters. The paper deals with a broad perspective of various load-balancing approaches done in the field by assuming the different performance metrics. The authors discuss that these approaches are multi-objective and there should be a good trade-off among these metrics to improve the performance.

Wireless sensor network have self-configured network which is consist of sensors and base station. The deployments of sensors are done in the required area to collect data and transmit it to the base station. The proposed energy-efficient routing protocol using threshold hierarchy for heterogeneous wireless sensor network (EEPH) protocol considers two parameters as distance and energy. The parameters are defined as the ratio of residual energy of current node to the total energy of all nodes. The distance is termed as the ratio of current node distance to the summation of distances of each node at that level from base station. The proposed protocol used three level of threshold hierarchy for different level of cluster-head selection process. The simulation result shows better performance and improved network lifetime compare to threshold-sensitive stable election protocol (TSEP).

The application of cooperative relaying is to limit the error in transmission and enhance the network throughput of nodes in the wireless sensor network. Intelligent relay selection scheme can be employed to improve the system performance further. This paper presents a performance comparison of proposed and existing relay selection schemes for cooperative wireless sensor networks. Conventional random relay selection scheme is compared with the proposed Fuzzy-based Relay Selection (FRS) scheme concerning frame error rate and network throughput. Simulation results depicted the effectiveness of the scheme providing better network throughput performance while limiting the error with improving the error rate performance in wireless sensor networks.

The rational factor generally discharges the exact requirements from the mindset to highlight the questions as it is with the actual cause of nature. Irrespectively, the irrational factors will flow from the mindset based on their individual scares and fearfulness of their subconscious thoughts of the mind. In general sequence, symptoms are the real facts to find out the cause for their solution to detect from the known values of attributes. Opinion mining is essential to find the data values and taxonomy to roll out the irrational factor on symptoms recognition. The research methodology is applied to find the new insight, which identifies the actual symptoms towards the disease belongings in a quick manner. The expected working hypothesis of outcomes may vary from the rational as dynamic issues against irrational, as unfounded factors, because the examination category depends on an expressive manner. The research work planned to expose and estimate roughly the rational and irrational factors of symptoms for the diseases. The paper presents features to detect the symptoms for the specific two diseases with positive or negative identification to correlate them in future survival.

E-learning is one of the modes of learning which is gaining popularity nowadays. To analyze the effectiveness of the E-class, there are many ways like feedback at the end of the lecture. But analyzing at the end will not serve the purpose. By analyzing the feedback of the listeners during the class improves the quality of the E-class. This can be achieved by analyzing the face and acoustics of the listeners. In a few cases, the data may not be obtained from both modalities. So, a novel method is proposed which will continue to perform and successfully recognize emotions even when one modality is absent. Product trends can be predicted by using this model for market analysis and ads can be improvised to reach their end target. Thereby, the manufacturers will receive relevant and better product feedback which leads to a rise in product sales. Studying how users view these multimodal posts will be of interest to sociologists and psychologists. The suggested project will provide a platform for publishing research papers in reputed journals, provides an opportunity for manpower training.

Every user wants to access their expected contents from legitimate web pages on the Internet by entering key phrases into search engines. Fortunately, crawler or spider (also known as web robots) retrieves the precise contents along with more number of vulnerable data, due to the crawler’s page navigations system. Search engines are simply bringing their responsive web pages based on assigned indexing values for all the visited links, without identifying hidden intrusion contents and web links on all the malicious pages. In the previous attacking era, spammers used (Spam) text for their attacks; later, many text-based filtering tools came into the picture to analyze spam text. Since images have very complicated features to extract its contents, after spam text identification, hackers nurtured their attacking methods based on the images, and they started to embed their spam contents into the image and spread it on victim’s web pages or email id. These images are called spam images. Images are scanned for its text extraction by Optical Character Recognition (OCR) system, and then, the mined contents are matched with spam text databases for spam text identifications. Based on threshold values of matched contents, spam images are identified and applied to the remedial actions. Subsequent failures of said techniques, hackers started to attack targeted victim’s data and system with the help of non-spam images. In which, they simply imbed their malicious web links and erroneous contents into the images, and those images are placed at the legitimate web pages in the form of some advertisements or stimulating user’s desires. Navigating to these intruded sites causes to DoS attacks, data and security breaching of a victim’s system. This paper is going to discuss web robots and its architecture, web content analysis and identification of intrusive substances of bogus images on web page contents.

The plant plays a vital role in human survival, so the preparation of the plant database and its identification system is essential for maintaining biodiversity. From the past two decades, plant identification based on leaf classification is on the main focus of researchers. It is proposed to build the convolutional neural network model for plant identification. However, overfitting is a significant issue in such a network. A regularization strategy like dropout is a good approach for addressing this issue. Along with this, transfer learning and parameter fine-tuning have come out with a good result on the same database. In the proposed system, performance of convolutional neural network is measured in terms of obtained validation accuracy.

Sentiment analysis aims to extract opinions, attitudes, as well as emotions from social media sites such as twitter. It has become a popular research area. The primary focus of the conventional way of sentiment analysis is on textual data. Twitter is the most renowned microblogging online networking site in which user posts updates related to different topics in the form of tweets. In this paper, a labeled dataset publicly available on Kaggle is used, and a comprehensive arrangement of pre-processing steps that make the tweets increasingly manageable to normal language handling strategies is structured. Since each example in the dataset is a pair of tweets and sentiment. So, supervised machine learning is used. In addition, sentiment analysis models based on naive Bayes, logistic regression, and support vector machine are proposed. The main intention is to break down sentiments all the more adequately. In twitter sentiment analysis, tweets are classified into positive sentiment and negative sentiment. This can be done using machine learning classifiers. Such classifiers will support a business, political parties, as well as analysts, etc., and so evaluate sentiments about them. By using training, data machine learning techniques correctly classify the tweets. So, this method doesn’t require a database of words, and in this manner, machine learning strategies are better and faster to perform sentiment analysis.

It is generally known that the skin infections are found in all the living organisms. Skin ailment is a specific sort of ailment introduced by either microorganisms or a disease. Out of the three essential kinds of skin malignant growth, namely Basal Cell Carcinoma (BCC), Squamous Cell Carcinoma (SCC), and Melanoma, the Melanoma is observed as one of the most hazardous in which endurance rate is extremely low. The early location of Melanoma can conceivably improve the endurance rates. Innovations in the skin disease recognition are extensively partitioned into four essential segments, viz., picture preprocessing that incorporates hair evacuation, de-clamor, honing, resize of the given skin picture followed by division. In this paper, a survey is carried out on the best in class in a PC helped analysis framework and further observes the ongoing practices in various strides of these frameworks. Measurements and results from the most significant and ongoing executions are broke down and announced. This research work has analyzed the presentation of late work that remains dependent on various boundaries like precision, dataset, computational time, shading space, AI procedure, and so on further the investigation carried out in this paper will help the scientists and researchers of the significant field.

In recent years, SMS spam messages are increasing exponentially due to the increase in mobile phone users. Also, there is a yearly increment in the volume of mobile phone spam. Filtering the spam message has become a key aspect. On the other side, machine learning has become an attractive research area and shown the capacity in data analysis. So, in this paper, two popular algorithms named Naive Bayes and support vector machine are applied to SMS data. The SMS dataset is considered from Kaggle resource. The detailed result analysis is presented. Accuracy of 96.19% and 98.79% is noticed for the chosen algorithms, respectively, for spam SMS detection.

A resonator is designed to operate as a filter resonating at 2.485 GHz using standard techniques. The design is initiated from the standard LC circuit. Using the insertion loss method, band pass or band stop second order Butterworth filters, with inductive and capacitive elements, are realized. On applying filter transformations to the $$\lambda /4$$ λ / 4 low pass filter, the standard $$\lambda /4$$ λ / 4 open and short circuit filter design is obtained. The open/short circuit realization of the filter using Kuroda’s identity was modified as closed loop resonators, with open and closed stubs, to achieve band stop and band pass operations. The optimization of the obtained circuit is carried out by adding stubs on to the section. A proof of concept of the operation of a resonator as a BPF for short circuit $$\lambda /2$$ λ / 2 lines and as BSF for open circuit $$\lambda /2$$ λ / 2 lines is carried out in this work. The miniaturization of the filter is done by adding open circuit stubs to BSF and short circuit stubs to BPF maintaining the electrical length of the design fixed. The further miniaturization of the design is carried out using standard techniques. The softwares used for designing and validation of the filters are ANSYS Designer and HFSS.

The wireless sensor network (WSN) is the most proven technology in today’s world. WSN has gained its applications that are different from other networks and even the usage of WSN has also inferred with other networks like VANETS that uses different types of sensors in its network. Arduino or raspberry pi uses different sensors to monitor and data can be obtained from any remote location at a very lower cost. Thus, miniaturization is possible using WSN. In this paper, WSN is being discussed in terms of evolution, scenario, hardware design, application, research issues, design constraints, research problem to provide an insight into WSN in different disciplines. Finally, the applications of WSN in agriculture with routing protocols are also being discussed. Thus, remote monitoring of agriculture using a sensor is demonstrated through this paper.

In the disease diagnosing process, identification of patterns is so crucial for the prediction of the disease accurately. A powerful disease prediction system is needed to detect the disease accurately by analyzing various parameters/symptoms and to make our systems learn from the past diagnosed disease and capable of adapting to new methods. To deliver the medical association for the analysis of syndrome among patients, a graphical user alliance will be refined. Doctors and medical practitioners can readily utilize the GUI as a screening tool. In this paper, various techniques available for disease prediction such as k-nearest neighbor (KNN), Naïve Bayes, support vector machine (SVM), logistic regression, and decision tree have been explained. A general review is done on the prevailing and anticipated models for human disease prediction and a reasonable study on these techniques based on quantitative dimensions such as detection rate (D-Rate), false alarm rate (FA-Rate) is carried out. Strategies have been proposed for diagnosing liver, heart, and diabetes illness in patients by utilizing machine learning procedures. The three machine learning procedures that were utilized incorporate SVM, logistic regression, and kNN. The framework was executed utilizing every model and their exhibition was assessed. Execution assessment depended on certain exhibition measurements. Finally, a model has been developed in which the person can insert his symptoms and the disease identification is done that can give the result based upon the inputs.

In the present world, the prediction of the results of football matches is being done by both machines and football experts. Football as a game produces a huge amount of statistical data about the players of the team; the matches played between the teams and the environment in which the match is being played. This statistical data can be exploited using various machine learning techniques to predict various information related to a particular football match, namely the result of a particular game, injury of a player, performance of a player in a particular match, and spotting new talents in the game, etc. In this work, previous works are reviewed on the prediction of the outcome of a football match, evaluate the merits and demerits of different approaches and then attempt to design a prediction system powered by Recurrent Neural Networks (RNNs) and Long Short Term Memory (LSTMs).

In recent years, healthcare applications have been gaining market in the research area. The objective of the work is based on mellitus diabetes and seeks to sort people in such a way that people who have affected most will be handled with high preference, moderate attack people will be the next level, and mild attack people will be given less priority. The methodology applied here is hesitant PROMOTHEE fuzzy in soft computing which helps to find the outranking and the best selection. The preference function is to be defined in PROMOTHEE to find the outranking decisions for the alternatives using criteria. This method is more powerful because preference can be given to the alternatives using their criteria. The rows are always treated as alternatives and columns are considered as criteria. Dominance and sub-dominance are the factors that play a vital role in the methodology described. The outcome of the research work focuses on giving the high peace of treatment to the most affected people and to safeguard their lives from the sudden mortality. The selection and the ranking have been given the foremost preference and priority in the research work with high precision of PROMOTHEE calculations in the healthcare application.

This paper aims to create a recommendation system for scientific articles. The system can help both researchers and students in finding suitable articles related to their fields of study in a time-efficient manner. This paper proposes a model that uses the DBLP-Citation-network V10 Dataset and a second model that uses a custom dataset created by combining an existing CORA dataset with web scraped information. The core idea of this paper is to use the latent relationships found in the citation data to find more relevant papers. An additional feature provided by this system is that it can scale to any number of new scientific articles. The proposed approach uses various pre-processing techniques and algorithms to calculate scores such as PageRank, TextRank, cosine similarity, author score, etc. Since the content-based and collaborative methods are used to calculate the scores, the proposed system is a hybrid recommendation system. The final recommendations are made by normalizing the scores and selecting the top 10 scores. The system was found to provide suitable recommendations to the user achieving accuracies of 76.4% and 81.4% for the two models, respectively.

The Evoked Potential (EP) is a term that refers to the response generated by the brain in effect to the external stimuli. The response is measured by the strength of the electric potential generated by the brain. The measured response varies depends upon the flickering speed of the stimuli through which the stimuli can be identified. In this study, the activity of the brain while perceiving the visual stimuli of varying frequency has been investigated. The study was made on the Electroencephalography (EEG) dataset designed by the authors that were acquired from the healthy seven subjects whose mean age is 22. The acquired is transformed to the time-frequency domain by applying wavelet transforms, and the statistical features were extracted from the acquired and transformed EEG signals for classification The proposed study applied machine learning algorithms to classify the appropriate stimuli. The study has experimented with machine learning algorithms like Support Vector Machines (SVM), random forest, K-nearest neighbour, multi-layer perceptron, linear discriminant analysis with the accuracy of 93.14%, 97.85%, 71.08%, 79.65%, 81.46% and 94.09%, 99.06%, 90.02%, 85.16%, 82.91% in time and time-frequency domain, respectively.

Nowadays, in the era of Covid-19, the demand for video data have grown exponentially due to online activities at every level. Online classes, conferences, meetings are everywhere all around us. All business activities are governed by online platforms. Due to the high demand for video data, it has been the main area of research to compress the video size. For all these activities, lossy compression is very useful, due to the limitation of the human visual system, as it can tolerate micro details in any image or video. Block matching algorithms are mostly used to determine motion estimation of the blocks for the video compression in available standards for the video encoding like AVC–Advanced video coding, HEVC–High-efficiency video coding, HDR video encoding compression (N-HVEC) and VVC–Versatile video coding. Motion estimation along with deep learning-based method is showing very motivating results in the field of video compression. Performances of block matching algorithms depend on the matching criterion. In this paper, various available matching criteria are reviewed based on—mean of the number of search locations for every block, average MAD per pixel and average peak signal to noise ratio (PSNR).

Autonomous vehicles, without the help of a human, support challenging tasks for sensing the environment and vehicle navigation. The driving behavior is controlled automatically from the observed surroundings using many supervised learning methods that provide action output based on matching the visual inputs and training labels. Most essentially, deep learning algorithms offer improved processing of observed input data but with the increased training, the complexity in processing the real-time data eventually becomes complex. In this paper, an autonomous driving model driven by a behavioral model is designed incorporating (a) recognition, (b) planning and (c) prediction modules. Each module is designed to regulate the processing of input trajectory video data. Additionally, deep learning classifiers are included to improve the automated ability of planning and prediction modules. Initially, the recognition module is planned to limit the redundant data from the raw input data. Secondly, the planning module is designed with convolutional neural network (CNN) to classify the predictable and unpredictable objects from the surrounding trajectories occurring in the line of sight. Finally, the prediction module is designed with recurrent neural network (RNN) to predict the future driving patterns based on the present condition and past driving outputs. The simulation results show that the proposed hybrid deep learning behavioral model offers improved autonomous driving than other existing autonomous driving models. The results of different environments prove that the proposed hybrid model offers increased scalability in terms of improved recall rate of 95.15%, 96.13% and 97.72% in terrain, dense and light traffic zones, respectively, than existing methods.

In video examination, image or video quality improvement is an emanant area of research. Images and videos acquired by the imaging sensors have rich and detailed information. The viewable representation of low-quality videos can be improved using video enhancement techniques. A video enhancement and reconstruction framework for facial images with low resolution are proposed in this paper. The video enhancement phase improves the appearance of images that are not sharply defined. In the reconstruction phase, the facial regions are selected from the high-quality video frames and the resolutions of these selected facial regions are increased. After reconstructing the facial regions, the aim is to authenticate and verify the human face through a face recognition network. A detailed analysis of the experiments is reported and is observed that the results obtained are significant.

Facial emotion recognition using expression analysis of human face is one of the important areas in video analysis which uses deep learning concepts to predict the state of mind of a person. Human facial expressions include means of communication through body language, eye behaviour and gestures. Facial emotion recognition discussed in this paper deals with the analysis of the state of mind of a person through his/her facial expressions. The main aim is to propose an unusual behaviour identification and alert generation system using facial expressions. The primary step in facial emotion analysis is to identify faces. After detecting the face, the expression analysis of the person is done by using a convolutional neural network. Through this analysis, unusual behaviour can be identified and an alert is sent to the corresponding authorities so that it can be useful for them in case if any incidents happen.

In recent times, digital data is widely used for processing and sometimes the processed data is needed to be transmitted from one organization to other. For such situations, secure transmission of digital data over the network is needed. Digital watermarking provides authority by preventing unauthorized access of multimedia data like audio, video and images. Our research work is based on an SVD-AES-based digital image watermarking scheme to provide secure transmission of digital images over networks. Initially, SVD-based watermarking is carried out. After watermarking, image is encrypted by means of an Advanced Encryption Standard (AES) algorithm. After that, encrypted image is shared over networks and finally AES decryption gives the watermarked image. The proposed method uses AES-128 encryption algorithm. The algorithm shows robust nature towards attacks like content removal, copy and paste attack and also cryptographic attacks like brute-force attack.

Power line communication (PLC) is a technology of transferring data over power lines. As the power lines were not designed to carry communication signals, noise in the form of impulses and fading corrupt the signals. Researchers have proposed various techniques to make communication over power lines feasible such as interleave division multiple access (IDMA), code division multiple access (CDMA), and orthogonal frequency division multiplexing (OFDM). The most promising results have been obtained by employing a combination OFDM-IDMA scheme which tends to resolve both the issues caused due to impulse noise and fading. But the key factor in the performance is based on the type of interleaver used in the system. In literature, many interleavers have been proposed such as random, prime, power, tree-based, and power rotational out of which the performance of power rotational interleaver has not been implemented in OFDM-IDMA over PLC. The paper aims to bring out the key aspects of power rotational interleaver and analyze the performance in terms of bit error rate by performing simulations in MATLAB and also present a comparative study with the other interleavers.

Indoor positioning and navigation systems are used to track entities in indoor spaces using Global Positioning System (GPS) and other satellite technology. The objective of this research is to demonstrate a semi-dynamic offline 3D indoor navigation setup using Received Signal Strength Indicator (RSSI) values of WiFi routers in the building with the support of an Android application. The data collection module in the application collects RSSI data along with the user’s co-ordinates (Online Phase) and a TensorFlow model is trained on the collected data, converted to TensorFlow Lite format and hosted online. In the Offline Phase, RSSI data is recorded, processed with the machine learning model in the device and passed to a module in the app developed with Unity that visualizes the user’s co-ordinates in a three-dimensional model of the building. This paper is a bare-bones implementation of the above mentioned ideologies of an indoor navigation system and can be fine-tuned to use in specific applications.

In defence applications, tracking of every individual is necessary. The devices and techniques used in the present world are manual and externally equipped. The automation of real-time tracking of every soldier without any external hindrance is the need of the hour. This task can be accomplished by designing an antenna with high frequency. The designed antenna is fabricated onto a textile (jeans) material which will be worn by the soldiers during training or emergency periods. The antenna is designed to operate at Wi-Fi frequency (2.4 GHz) for simplicity of sensor (GPS) integration and utility. It is interfaced with the Arduino board which acts like a transceiver to send the real-time location of the soldier in the form of latitudes and longitudes. This setup is completely automated and integrated on the clothing of the soldier, which eliminates carrying any extra device for tracking himself. The designed meander line patch antenna on the ‘jeans’ textile material has the specifications which are optimised to produce maximum directivity and gain despite the human torso. Serial communication is used to transmit and receive sensor data via a pre-coded Arduino board. The project can also be extended to several other applications such as heart monitoring of soldiers, etc.

The advancement of internet technologies has led to the appearance of cloud robotics. Besides the numerous benefits of cloud robotics, security threats affect the growth of this field. Therefore, a secure and efficient authentication mechanism are needed to prohibit numbers of outsider and insider attacks present in the system model. Authentication of the robots plays an important role in safely access remote resources and remove the many vulnerabilities present in the network. Elliptic curve cryptography (ECC) gives better results in scarce computing resources and limited energy systems such as cloud robotics. In this paper, an ECC-based authenticating scheme has been proposed to facilitate access to cloud-based robots. The proposed scheme uniquely identifies the robots before using cloud-based robotics services. The security analysis using the proverif tool establishes that the proposed scheme resists many well-known attacks.

Web of thing is modern technology and a subset of the internet of things (IoT) which brings many application and new possibilities to improve usability and the interoperability of IoT. WoT has many challenges spatially data analytics and storage due to the increasing number of sensing devices capable of acquiring huge amounts of data. This paper presents two techniques of analytics sensors data which distributed in smart home to measure temperature and humidity of rooms: firstly, clustering analysis to sensors measurements of room’s temperature and humidity in the smart home as one solution to the challenge of data analytics and storage and to determine the temperature and humidity patterns which help in making better decisions at right time, secondly, prediction model of room’s temperature and humidity as a safety system in the smart home when compared the current value with predictive value depends on historical data to detect the deviation of the measured data from the sensors, then the output of two techniques are shown in designed web pages to provide better services for the citizens living in the home. Those techniques are implemented by using intelligence neural networks and fuzzy logic, fuzzy adaptive resonance theory neural network (Fuzzy ART—NN) for clustering model and long-short term memory recurrent neural network (LSTM—RNN) for the prediction model. The performance of the clustering model evaluated by training time and the accuracy when the results are shown the clustering approach has short training time and accuracy reached 99%. While the performance of prediction approach evaluated using root mean square error (RMSE) and training time, and the results are shown RMSE reached 0.02 and training time approximation of 4.76 s for 4550 samples.