ICDSMLA 2021

This main objective of this paper is to develop a detection system using Arduino, GPS, GSM and MEMS when a vehicle met with an accident. Here, MEMS technology detects the sudden change in the axes of vehicle and GSM module will send a message with the location of the accident to nearby hospital and relatives to alert them. GPS module is used to share the location of accident in the form of Google map link, derived from the latitude and longitude form. The alert message received by other side also contains the speed of vehicle in kmph. The proposed idea can also be used to track a vehicle continuously, by just making few changes in both software and hardware. The alert system has been developed in real time and performance is discussed.

To manage attendance of each student is much more complex when it’s done manually, and it has to be done for each class. To overcome this, we can use face recognition application where it takes less time comparatively, and it is stored in the database; this avoids confusions, proxies, and error of storing data when done manually. The students need to fill their data along with parent’s details, so when the students are absent, the message will be passed to them. After taking attendance through the application, mails and message of the data entry of attendance will be sent to the respective teachers. In advance, the details need to be entered beforehand, so when the students come in front of the camera, their face is recognized by comparing them from database containing faces. When it is unable to recognize, the faces will be stored in unknown, so when there is a mishap of being absent, it can be checked later on. This method of attendance is much more successful. Compare to other algorithms, haar cascade and local binary pattern histogram is best due to their robustness and less false rate.

Document data is captured through optical scanning or digital video, resulting in a file of picture elements, or pixels, which serves as the raw input for document analysis. These pixels are samples of intensity values taken in a grid pattern throughout the document page, with intensity values ranging from OFF (0) to ON (1) for binary pictures, 0–255 for gray-scale images, and 3 channels of 0–255 color values for color images. The initial stage in document analysis is to process this image so that it may be analyzed further. Thresholding is used to convert a gray-scale or color image to a binary image, and noise reduction is used to remove superfluous data. The goal of this paper is to summarize some thresholding technique for image processing.

Finite Set-Model Predictive Control (FS-MPC) based torque and flux control (popularly known as PTC) is one of the advanced control techniques used for the adjustable speed operation of industrial drives. This control technique offers fast dynamic response and retains the decoupled nature present in the conventional counterpart. One of the main challenges of this control approach is to select the suitable weighting factor (WF) for the respective objectives in cost function. In general, a heuristically selected WF is assigned to the flux objective in the cost function. The selected WF is constant irrespective of the drive operating conditions. In this paper, dynamically weighted objectives with modified cost function are introduced to overcome the above problem. To demonstrate the efficacy of proposed dynamic weighting approach, simulation results are illustrated for an induction motor (IM) drive with different operating conditions.

Nowadays, every individual is familiar with the COVID-19 pandemic which has caused great turmoil in everyone’s life. Also, they are aware that there is no medicine or drug to cure COVID immediately, and people are at the risk of losing their lives. Lack of vaccines or delay in vaccine production for mass results social distancing being the only measure to tackle this pandemic. As a result, social distancing has proven to be a very reliable and efficient way to diminish the growth of this disease; the reason why lockdowns are imposed, and people are asked to keep some distance from each other, for their safety as there will be minimal physical contact. Machine learning and artificial intelligence come into the picture in every solution to a generic problem the community faces nowadays like in medical, supply chain management, face detection, etc. Using the power of AI algorithms, the paper aims to develop a robust system to monitor and analyze social distance measurement protocols at public places during the COVID-19 pandemic with the help of CCTV feed and check whether they abide by the safety protocols or not by measuring the distance between them. The proposed approach is implemented to enumerate the number of violations at a popular public place to prevent massive crowds at particular periods. The proposed method is suitable to construct a scrutiny system at a public place to alert people and eschew mass gatherings that can be concluded using achieved results. The paper also has an analysis of the performance of different models of R-CNN, Fast R-CNN, and YOLO. YOLO architectures are validated based on object detection and object tracking rate in real time.

Binarized ReLU activations are considered as a metric space equipped with the Hamming distance. While for two-layer ReLU networks with random Gaussian weights it can be shown theoretically that local metric properties are approximately preserved, we experimentally study the discrimination capability in this Hamming space for deeper ReLU networks and look also at the non-local behavior. It turns out that the discrimination capability is approximately preserved as expected.

Machine learning task is broadly divided into supervised and unsupervised approaches. In supervised learning, output is already known and we have to train the model by giving lot of data called labeled dataset to train our model. The main goal is to predict the outcome. It includes regression and classification problem. In unsupervised learning, no output mapping with input as well as it is independent in nature. The dataset used in unsupervised machine learning is unlabeled. The main focus of this paper is to give detailed understanding of supervised and unsupervised machine learning algorithm with pseudocodes.

In the past 10-years, from 2008 to 2018, the annual sort of skin cancer cases has raised by fifty-three percent due to increased ultraviolet exposure. Though skin cancer is one of the foremost deadly variants of malignant neoplastic disease, a faster identification can cause a very high chance of survival. The primary step of diagnosis of a lesion by a specialist is visual examination of the suspicious skin lesion. It is found that a specialized doctor who treats skin typically carries out a sequence of phases, initial from eye examination of distrusted injuries, followed by dermoscopy (magnifying injuries microscopically) and later with a diagnostic test such as biopsy. This process is time consuming and the patient might progress to future stages. What is more correct designation is subjective; most effective skin doctor has associative accuracy of eightieth in properly diagnosing the carcinoma type. Adding to those difficulties, there do not seem to be several masterful dermatologists out there for public aid. In association, correct diagnosis is important, adding to the similarities of some lesion types; what is more important is that the diagnostic accuracy correlates powerfully with the masterful experience of the medical. An increased help to the skin doctor is delivered through the emerging technologies of deep learning. The basic goal of this method is to train a model to solve the problem by reviewing cancer images. The model will be constructed without any programming skills, which is a unique quality of the presentation. Convolutional neural network (CNN) primarily based classifiers became the most effective selection for cancer detection within the recent era. The analysis has indicated that classifiers that supported CNN classify carcinoma pictures similar to dermatologists that has allowed a fast and life-saving diagnosis.

Agriculture provides a living for around 58% of India’s population. Agriculture, forestry, and fisheries were expected to generate ₹19.48 lakh crore in FY20. Given the significance of agriculture in India, farmers might benefit from early forecasting of agricultural yields. The study focuses on predicting agricultural yield, for Karnataka state using the regression with neural network model. The final constructed dataset takes parameters like agricultural area, crop, taluka, year, season, district wise annual rainfall (mm), district wise maximum and minimum temperature (°C) and harvest or yield for the time period of 1997–2017. The underlying model is built utilizing a Multilayer Perceptron Neural Network, a ReLu Activation function, an Adam Optimizer, and 50 epochs with a batch size of 200. The end of the training gained 96.43% accuracy on test data. Several additional well-known regression algorithms such as Multinomial Linear Regression, Random Forest Regression, and Support Vector Machine are also constructed and trained using the same dataset so as to compare their performance to the base model. From the final comparison results it was found that neural network model has outperformed classic machine models for crop yield prediction in terms of both mean absolute error and accuracy.

The rapid growth of social media sites in recent times has introduced a special environment for researching human actions. Micro blogging website (Twitter) which also allows users to access and communicate their opinions on various of topics, occurrences, brands, and service providers. Tweets have been classified into different subgroups relating to the subject of the research. Various machine learning algorithms, such as baseline, Naive Bayes classifier, support vector machine (SVM), and many others, are presently used this to categorize posts on Twitter into favourable and unfavourable groups based on their views and opinions. This paper describes a solution to improve comparison between Twitter accounts besides trying to implement Naive Bayes using perception vibrant data for training from of the Twitter database. SentiWordNet, in combined effect with Naïve Bayes classifier, can help increase tweet accuracy rate by providing hope and optimism, hatefulness, and integrity ratings for phrases in Twitter messages. Tweepy, which is a Python package and Python-Twitter APIs are used in the actual introduction of the new system. The main work of this paper is to perform real-time streaming, collect the data set while performing real time and make comparison between them using Naïve Bayes.

As the technology is developing day by day, shopping malls should be capable of handling the crowd smartly. The various items are purchased in shopping mall or markets with the aid of shopping trolley. After purchasing, the customer needs to pay the bill in the counter. In that case, they have to wait for long time for the person in billing section to scan each and every product. This is why there has been an inevitable demand for quick and easy payment of bills in shopping places. This project is developed for the main purpose of saving the customers time and makes shopping much easier. This is based on Raspberry with a LCD and QR or barcode scanner and a wireless technology called Wi-Fi. The LCD used is a 20 × 4 and Wi-Fi modules make the wireless network to work easily between a certain ranges. A brief description about the project is it produces a method wherein the customer can scan any product using the barcode scanner installed in trolley and the information is displayed on the LCD screen. The system is provided with four options such as add, multiple, delete, and stop button which can be used in the different case. The customer can scan any number of products, the cost is added to the final bill, and the customer gets the bill to their mail id given.

In recent days, activities are completely automated and are more flexible because of Internet of Things [IoT]. Due to IoT devices, effective integrated communication between all types of embedded devices is possible. The devices should have communication in stipulated range through the support of Internet. This significant functionality is built by the IoT protocols. In the digital world, internet-connected devices has enabled the IoT protocols to be one of the significant fields in computing, and it acts as a bridge between the physical and the cyber-world. This paper gives a brief overview of Internet of Things (IoT) and standards used to it. Important IoT protocols and their review are presented. The similarities and dissimilarities of Advanced Message Queuing Protocol (AMQP) and Message Queue Telemetry Transport (MQTT) protocols are also discussed. The main aim of this review papers is to provide a functional view of IoT architecture, standards and protocols.

Nowadays users can store the information remotely in the cloud and whenever they want, they can access information at any place with distributed storage system. For the information stored in remote places, information respectability reviewing is used to ensure truthfulness of the information put away in the virtual storage. In distributed virtual storage, for example, the Bank Transaction Records system, the virtual storage may consist of important information. The important information is not to be displayed to the unknown persons while the virtual records are accessed. Accessed documents are encoded and that can understand the sensitive data carefully and neatly. The most effective method to acknowledge information offering to sensitive data are carefully handled in remote places, data reviewing not been systematically examined till today. So, to address this issue, we are bringing the far-off information examine to ensure the idea that acknowledges information directed to sensitive data handled carefully here. In this plan, a tool is brought to clean the information squares comparing to most wanted data of the record and changes these information squares’ marks to essential one for the cleaned document. These marks are carried out to check the completeness of the cleaned document in the span of reviewing. Afterward our scheme uses the record stored in the virtual storage that is accessed and used by needy ones based on the precondition that the important content is secured. The distant information rectitude inspect is ready to be competently executed. Suggested plan is based on personality-based-cryptography.

This study proposes to classify waste as recyclable and helps them to improve the sorting process of compact waste collected from the public. The proposed system is an advance hybrid multi-layered deep learning system which uses an algorithm of CNN for capturing the image of waste and sorting the compact waste which can be further recyclable to a useful product. It reduces the man efforts and helps the mankind to automate the system of garbage separation. This projected system is using the CNN model which is trained and assessed by niche technology to collect and sort the waste. It is proven that the speed of detection the waste is above 90%, which is quit precision relying on image only inputs.

The aim is to develop a detailed and precise face recognition-based home automation system which is based on IoT using modules and microcontroller for wireless security. This project mainly highlights and focuses on controlling certain devices such as home appliance by integrating it to Internet and making it an IoT-based system, and building an accurate and smart wireless home security system using face recognition system and Wi-Fi as communication protocol and certain network protocols. Looking at the current scenario, we find an existing system which is assimilated with home automation system. The paper is being focused more on reducing the limitation with respect to being more widespread in terms of usage. As far as the study area of this paper is considered, NodeMCU microcontroller unit along with relays module is used to control electrical appliances. Face recognition module will provide smart security wherein a captured image is sent through an E-mail/MMS to the owner using IoT-based cloud service such as Internet when a face is detected. User can be authorized to control security controllers using certain application after authenticating. Hence, resulting in being useful to people with physically challenging disabilities.

Providing computer security is one of a significant challenge. Software and its mechanisms have been established to provide the security to avoid intrusion, which includes Intrusion Detection Systems (IDS). IDS helps to detect the exertions to outbreak a network and detect anomalous actions and its activities. It includes the details of uncertainty in probing of different kinds of attacks. IDS demands the need for combination of Machine Learning (ML) methods integrated into a hybrid model. In this paper, the Hybrid Machine Learning (HML) model is proposed that predicts the attacks against the network provided with better performance. The proposed system includes an innovative IDS having good network act helping to perceive the strange attacks, achieved by ML algorithms such as Decision Tree (DT), Naive Bayes (NB), Random Forest (FR), K-Means, and SVM. The algorithms with the good results are used to build a hybrid model. The proposed HML method improves the accurateness and efficiency for detecting the attacks by the IDS system. This research work recommends a system for access within a hybrid model, i.e., Hybrid-Network Intrusion Detection (H-NID) which is a hybrid network combination of DT, K-NN, NB, SVM, and RF to extract temporary and local data of network traffic which advances the accurateness of IDS. The training phase in H-NID uses stage wise approaches to scale out the model. This technique decreases this consequence quantity of unparalleled trials of various attacks on training of model execution. It progresses the strength of training and prediction. Lastly, test of H-NID is done for several types of network traffic from the CICIDS-2017 database as it works on a real network traffic dataset that simulates real-world conditions. The predicted results demonstrate that H-NID got 98.50% of accuracy, and the accuracy for each type of attack remained more than 94.65%, which achieved excellent results in all models. Various rules and restrictions do not work well.

Gait analysis is a behavioral biometric that classifies human, based on how they walk and other variables involved in the forward movement. In this study, we have attempted to comprehend the significance of the upper portion of the body in gait analysis for human recognition. The data for this study came from the CASIA dataset, which was donated by the Chinese Academy of Sciences’ Institute of Automation. We began by extracting the gait energy image (GEI) from the dataset and employing principal component analysis to minimize the dimensionality (PCA). For classification, random forest, support vector machine (SVM), and convolution neural network (CNN) algorithms are implemented to recognize the human subjects. This paper provides experimental results to show the accuracy attained when classification is done on GEI of full-body images is higher than the accuracy attained when classification is done on GEI of the lower portion of the body only. It also shows the significance of the GEI of the upper portion of the body.

Wireless mesh networks have risen to prominence as a critical component of wireless communications. The mesh network topology is the interconnectedness of all nodes in the cluster. Endpoints, users, gateways, bridges, and other equipment make up a network. Mesh nodes are typically less adaptable than other systems because reconfiguration is easier to foresee, resulting in delays in transmitting data. The basic architecture of a mesh network consists of the number of wireless nodes connected via base stations. The mesh network has different client-based structures; infrastructure and hybrid mesh structures. There are several advantages of a mesh network, such as even with a vast proportion of devices in the system, a WMN provided efficient results. There are various applications of the WMN. For example, the mesh network is used in smart meters. Electrical smart meters have been installed in homes to transmit readings from one place to another to reduce personnel. The installation of WMN is easy to set up and deactivate, making the network highly versatile with fewer or more units. This study examines several routing protocols which are used in WMN as well as evaluates their functionality.

Nowadays, sound serves as a crucial factor in all facets of human life. Staring from automating personal security systems to critical surveillance systems, sound is an indispensable component. The practical implementation of the present day automatic sound recognition systems in real-life settings is inadmissible due to their poor detection accuracy. However, deep learning-based systems overcome the incompetence of the traditional machine learning-based models, and it can be used to develop automatic sound classification systems. This work proposes a deep meta-model for categorizing environmental sounds on the basis of the spectrogram images generated from these sounds. In the proposed approach, spectrogram images of environmental sounds are used to train five different deep learning models, and the predictions from these base models are then stacked using the proposed deep meta-model. Experimental results on two benchmark datasets such as ESC-50 and UrbanSound 8K demonstrate the fact that the proposed deep meta-model is a promising alternative to the conventional approaches for environmental sound recognition.

In the era of spatial computing, humans are transitioning from how communication was done in the past with rigid devices to how interaction will be done with today’s dynamic and interactive technology. It has changed the world in the terms of communication, understanding and providing solutions to complex problems. A computer is no longer just a gigantic and static machine. Computing devices can be used in anything from wearable devices to smart assistants on your computer. Via spatial computing, everyone can build machines that wander the streets, trying to fix violence individually. While our efforts to make technologies more interactive, which are often not as effective as it would like. Spatial computing plays a vital role in area of medical science, industry and automation industry as well as is spreading into every aspect of everyday life currently. More significantly, spatial computing’s most useful characteristics inform us that this innovation is not just entertainment and games but much more.

The optimal water conservation in forming using IOT-based WSNs using cloud accessing is aimed to benefit the millions of Indian farmers toward their farms timely watering requirement using scientific methods and the state-of-the-art technologies with optimum usage of resources. The system consists of wireless ambient monitoring sensors like soil moisture, temperature, etc., combined with low power wireless networks. The wireless sensor network acquires the data on continuous basis (24 * 7 mode) from the sensors and analyze in conjunction with the Indian Metrological Department’s weather predictions and makes intelligent decisions based on the available data and executes the watering activities. The information about the execution of activities will be sent to the user through email/SMS using Internet. The entire sensor’s data will be logged and stored in the cloud, and this will be used as a base for fine tuning of the intelligent irrigation systems for future to effectively meet the agricultural demands. Excess watering causes loss of minerals and rooting up of the plants and wastage of resources. This system conserves the usage of man power, electricity, and water resources optimally. In this paper, the system allows the user to schedule the irrigation timings with a provision to override the timings from any computer with a web browser. No special software or clients are required, just a web browser on a computer or mobile phone.

While building convolutional network-based systems, the toll it takes to train the network is something that cannot be ignored. In cases where we need to append additional capabilities to the existing model, the attention immediately goes toward retraining techniques. In this paper, I show how to leverage knowledge about the dataset to append the class faster while maintaining the speed of inference as well as the accuracies; while reducing the amount of time and data required. The method can extend a class in the existing object detection model in 1/10th of the time compared to the other existing methods. S-Extension patch not only offers faster training but also speed and ease of adaptation, as it can be appended to any existing system, given it fulfills the similarity threshold condition.

The present COVID attack has significantly accentuated vulnerability of infections at community level where their basic healthcare and immunization program are being implemented by the ICDS scheme launched in 1975 through a slew of AWCs scattered all over the country. The AWC acts as a primary health center that provides supplementary nutrition to children (between 0 and 6 years of age) and pregnant and lactating mothers besides providing preschool education to children in the age group of 4–6 years. The ASHA workers associated with AWC under then NRHM are the first hand health workers available to them at the community level who act as a bridge between the dispensaries and the community members. With the outbreak of COVID-19, the role of ASHA workers has assumed increased salience as the governments are relying on them for community level combating of this outbreak. This paper takes a close look at fund allocations to the public healthcare sector among the developing and developed countries and also the interstate allocations and allocations for major schemes and the resultant impact on HDI.

To ensure stable operation of the grid-connected system even under high renewable energy penetration, this paper analyzes the phenomenon of fault ride through (FRT) using reactive power injection approach. The proposed approach is developed by limiting the current overshoot during the transient and time variant variations. This maintains a constant average active power and stabilizes the system to operate according to the grid standards. Further, numerical simulations and experiment are carried out by evaluating a symmetrical fault on a 16 kW three-phase grid-connected PV system to illustrate the performance of proposed method. The developed system regulates the DC link voltage, limits the maximum inverter current, and improves the voltage profile by injecting required reactive power during a fault and achieves dynamic grid support requirement.

In the era of smart homes, providing a smarter solution to water quality management is crucial, as water quality is being deteriorated over time. The proposed system uses Raspberry pi, Arduino Uno along with various sensors like turbidity sensor, pH sensor, TDS sensor to measure the quality of water. An ultrasonic sensor is used to measure the water level in the tank. A tank cleaning mechanism is proposed to clean the interior walls of the tank, thus preventing the growth of algae and bacteria. The system’s algorithm works efficiently to reduce power consumption, without hurting the functionality, so that the system can run continuously. Internet of Things (IoT) is deployed in the system for the user to control and interact with the system remotely using a mobile application.

Big datasets are characterized by large dimension consisting of hundreds of thousands features with uncertainties and imprecisions. It becomes a challenging task to represent these datasets in memory short environments. Feature selection techniques enable dimensionality reduction of such datasets by finding subsets of relevant features from the original feature space. To produce optimal feature subset, an ideal feature selection technique should be capable of handling the interdependencies and uncertainties in the features. In this paper, we propose a new hybrid feature selection technique called anchor-based fuzzy rough feature selection (ABFRFS) based on anchor graph-based learning and fuzzy rough feature selection for text categorization. Although anchor graph-based feature selection and fuzzy rough feature selection have been proposed independently earlier, yet the hybrid of anchor graph and fuzzy rough feature selection called ABFRFS is proposed with the intuition to overcome the inherent problem of representing big datasets in memory short environments and at the same time retain the interdependency uncertainty among its features while maintaining the clustering accuracy. It is observed that with the proposed ABFRFS technique the feature space is reduced to an extent of 98% on an average on the considered benchmark datasets with acceptable degree of clustering accuracy.

In the present-day scenario, we wear a variety of clothing but never know its fabric until unless a person touches or feels it. But is there any possibility to find its fabric without touching it? The answer for this query is yes, and we can find a fabric type of clothing without any touch. Here comes an application of IoT, where each problem can be solved by things in real-world problems. Fabric variation is a factor of recognition for a sensor where its resonance is the type and gives the sensor feedback. This type of mechanism scenario is used for regular period identifications of clothing. Pattern analysis in fabric can speak a matter a lot in differentiation of types in it.

Cardiovascular system plays a vital role in maintaining human health, and therefore, it is important to continuously monitor the cardiac activity. Electrocardiogram is the widely used technique to monitor the functioning of heart but it cannot be used at any place or at any time and it requires a clinical person for assistance. On the other hand, since cardiac activity is proportional to blood flow, determination of blood volume and its flow rate can be determined prior to ECG to know the cardiac function. One such method to determine this parameter is photoplethysmogram (PPG). It determines the volume of blood based on the absorption of light by oxy and de-oxy hemoglobin. Pulse oximeter is a device that works based on the principle of PPG. Even though it has many advantages, it also requires assistance and person feel discomfort on wearing this device during recording. Hence, image processing-based methodology has been presented with which a person can monitor his own cardiac activity at any place or at any time without any secondary assistance. Every common person possess a smartphone, therefore, integration of a technique to determine cardiac activity into smartphone will be advantageous to all persons. In this research, a methodology is presented that allows a person to monitor his own heart activity using only his smartphone and without any additional gadgets.

Earthquake warning systems are dependent on reliable and timely detection of p-phase arrival. However, the quality of early warning is achieved when the alarm is associated accurately with the probable damaging effect of the impending earthquake at a warning location. Hence, early prediction of peak ground acceleration (PGA) is necessary for real-time hazard assessment, automated preventive shutdown of safety–critical systems, shake-map generation for post-earthquake mitigation action, etc. In this article, a multilayer perceptron (MLP) neural network is trained for estimating on-site PGA using initial three-component p-wave features. Logarithmic transformation of PGA yielded an improved coefficient of determination (R2) from 62.09 to 70.15% when the MLP model is trained using features extracted from a 1-s-long accelerogram sampled from 0 to 7 s. At the same time, MAE and MSE reduced from 6.85 and 286.97 to 0.41 and 0.27, respectively. Effect of delay in feature window sampling from p-wave arrival further showed best R2 of 77.74% in case of 5–6 s delayed feature windows among 0–7 s.

Wireless communication has rapidly expanded value in recent years, making it the de facto commodity for both humans and machines. Today, wireless communication seems to make its way through almost anything that is being implemented. Therefore, finding efficient methods of this type of communication seems to increase every day. LoRa is one such technology in recent years and it has been gaining its importance ever since. In this paper, we summarize the analysis of LoRa specifically targeted toward the activities of a smart campus.

Big data are enormous amounts of information that may accomplish marvels. This has been a topic of special interest over the past two decades due to its tremendous development. In the medical sector, a different large data resource include medical files, patient hospital information, and clinical exam reports and Internet-based equipment. Data visualization may utilize the human vision system fully to lead people throughout statistical analyses and intuitively and easily relay the message buried while behind the data. The Big Data Medical Care Platform is designed based on the system needs and operational position, from collecting and exchange of health information, analysis, and knowledge management level. In this study, the VAE utilized to visualize big data for the administration of large data in the health sector is analyzed. Experimental findings show that in comparison to traditional methods, VAE can provide outstanding prediction performance, along with PCA speedily ICA, feature analysis, nonnegative matrix factor, latent Dirich allocation in terms of AUROC and accuracy.

In the previous decade, many studies have been proposed on intrusion detection systems that leverage machine learning techniques for attack detection. The majority of research employs manually derived features. However, this approach is time consuming, and lot of information is lost from the original data leading to inaccurate results. A neural network of hybrid convolution and long short-term memory is proposed to detect intrusions using the CICIDS dataset. CNN is used to extract spatial features, and LSTM is used to extract temporal aspects of traffic network data, thus providing an intrusion detection strategy. Compared to KDDCup99 dataset, CICIDS2017 dataset is the latest and benchmark dataset that includes all the recent cyberattacks. The proposed framework has an overall accuracy of 99.45% and F1 score of 99.4%, with each attack type having an accuracy of above 99.50%.

Virtualization is considered one of the green information technology (IT)s that will help lower the cost of technology and upkeep. Virtualization and multitasking operating systems have similar working capabilities. Virtualization techniques are now used in data centers to abstract physical networks, create large pools of logical resources consisting of central processing units (CPUs), memory, file storage, disks, networking and applications and distribute those resources to users or customers in the form of agile, modular and consolidated virtual machines. Cloud computing provides more convenience for any computing needs than conventional, offering custom virtual machine (VM). The proposed work focuses on two forms of tests in a preliminary analysis to assess the efficiency of cloud infrastructure. The first test is to assess network efficiency and the second to measure the performance of cloud computation. Here the focus is on the performance efficiency pillar of the Amazon web service (AWS) Well-Architected Framework. AWS is Amazon’s extensive and developing cloud computing network, including Platform as a Service (PaaS), and Software as a Service (SaaS) and Infrastructure as a Service (IaaS). The factors examined in this study include uptime, transfer rate, jitter and latency while transferring a file from the desktop to an Elastic Compute Cloud (EC2) instance. The findings demonstrate that VMs from a basic cloud computing infrastructure that are utilized to produce video have a significantly longer processing time than the same specification personal computer (PC). The network side has been considered a gateway to the success of degradation rendering in cloud computing.

From the last few years, load flow analysis of distribution networks has always been a focus of research. Many unique methodologies and methods for load flow analysis within the network have been developed, and simulation is being utilized to work out the network’s various features. This study discusses how network reconfiguration techniques aid in the reconfiguration of distribution networks due to excessive power losses, as well as how algorithm-based network reconfiguration processes aid in load flow calculations with DG placement.

Coronavirus disease (COVID-19) is a viral contagious disease caused by a newly discovered coronavirus. The COVID-19 virus primarily spreads from an infected person through droplets of saliva or nasal discharge when the person coughs or sneezes, and most people who have been infected with the virus usually experience mild to severe respiratory illness, and they recover with minimal or no treatment. COVID-19 causes mild illness in the majority of patients although it can be fatal in rare cases. Our project focuses on using an SPO2 level monitor and thermal scanning to monitor patient health and take precautions to avoid constant transmission, as well as providing support to patients by assisting them with basic needs with the help of food delivery agencies and non-governmental organizations (NGOs) and assisting with prevention. We use an enhanced version of the SIR epidemic model, which is further explained in this work as an IoT-based system which is being used for automated health monitoring and surveillance, this work aims to reveal certain facts about the current situation that are not presented by data, as well as predict and forecast future situations. AI-assisted sensors can be of major help to foresee whether or not someone is tested positive for the virus supported on indicators like body temperature, coughing patterns, and blood oxygen levels. The ability to track people's locations is another helpful function. All these problems collectively checked will make an efficient model to curb the virus.

There might be a lot of times that the user needs to know what he/she might have been going down with and keen to know what kind of disease they might be having so it is really easy for the user to contact the required doctor. The health disease prediction application comes in handy at many times. A lot of applications out there are concerned with a particular disease or some set of symptoms. Here, we propose a system that solves all the issues mentioned above. It is a mobile application where the user is instructed to add his/her symptoms and the disease that they might be having or tends to have will be displayed on the user screen. The user can access the latest news related to the medical. Here, we tend to use the machine learning algorithm that uses the trained data for predicting the disease. The user interface is built on JavaScript language using react native framework. Once the prediction is done using the major machine learning models, the result is being displayed to the user on the mobile application.

For years, universities have been using traditional methods as they bring about a sense of familiarity and comfort. However, the key to increase productivity and enhance the learning experience lies in the modernization of our college campuses. Our work focuses on two of the most pressing issues—attendance and unsolicited energy consumption. Our goal was to overhaul the manual system, without compromising its integrity. We have developed a portable RFID fingerprint scanner that uses capacitance to gauge the depth of the finger and collect the fingerprint that must be activated by the professor, thereby making our system fail proof. The data generated is stored in a cloud server and is used to draw behavioral patterns. We are accustomed to leaving our fans and lights unattended, which poses a serious environmental threat. Our solution is to build a smart eco-system to detect the presence of a person and automatically switch on the necessary appliances. Further, this in combination with various algorithms can be used to adjust the speed of fans and brightness of lights based on external factors.

Liquid level monitoring and control is the prime objective in all real-time industrial applications and also in domestic applications. Water level control in power plants is of more significant when contrasted with other parameter control and monitoring. The prime intention of this work is to propose a mode for effectively controlling a liquid level in a tank utilizing fuzzy logic control and IoT technologies. The controller values are premeditated and simulated using MATLAB platform which utilizes Mamdani type Fuzzy implication System. Initially a PI controller is refined to compute the complete working values of controller attributes. Then, using the specified variety of measures, a fuzzy logic control scheme (FLC) is devised for optimal Kp and Ki which adapts itself to the variations in error. The controller receives two inputs: the error input between the reference and controlled outputs, and the latter is the derivative of the error. The valve position is considered as the output. The setup is emulated using MATLAB –Simulink to assure that the controller functions properly in the presence and absence of disturbances. When contrasted with the traditional methods the FLC shows enhanced performance in the aspects of reduced overshoot and reduced settling time. The entire setup is interlinked with Texas processor for real-time IoT monitoring.

In major and rapidly developing urban regions across the world, traffic congestion is unavoidable. Congestion during rush hour is an unavoidable consequence, especially in large metropolitan. In the current circumstance, the traditional technique works well only if the count is low, as the density of vehicles on one side of the lane road grows if the traffic is heavier on one side of the road than on the other; otherwise, the approach fails. As a result, our goal is to create a traffic system that can switch signals as well as track and handle signals in real time. Signal switching will be done in this project based on real-time image detection and accuracy of vehicular traffic. Another goal is to put in place a traffic management system that gives priority to the lane where the ambulance arrives that offers ambulance detection.

Animatronics is a technology where machines/robots mimic human and animal activities like walking, arm movements, facial expressions, etc. The main idea of this article is to design an animatronic hand that operates on the information received from the gloved hand including an infrared sensor. The data received from the infrared sensor is sent to the servo motor, which rotates at a certain angle. Data transmission between the animatronic hand and the gloved hand is completely wireless using the CC2500 module. This can be used primarily in applications such as pick and place and where there is a risk of human intervention.

Coffee production is the main agricultural activity in South India. Karnataka, as the centre of origin for Coffea Robusta and Arabica, hosts a large diversity of germplasm. Coffee diseases in leaves have become major threat to the production and result in commercial losses in agronomic industry. It is important for coffee growers to be aware of the damage caused by the diseases and identify them properly in the initial stage itself to take appropriate measures to prevent the crop loss and to keep the coffee plants healthy. In this coffee research, the authors summarize incidences and distributions of the variety of coffee diseases, most importance of coffee and recent techniques for detection of coffee plant diseases. However, the methods projected are generally incomplete in scope and rely on capture environments in demand to work accurately. The shortfall of important technology would partially describe with tough challenges due to various factors like complex environment that cannot be simply detached from area of interest (leaf and stem). Symptom edges would not be demarcated properly. Analysis of images would be complex task with unbounded capture conditions. Symptoms with a broad range of idiosyncrasy may produce disease. The symptoms bring out various diseases that may look like similar, and they might overfit together. Here, provides a study and examine of above challenges, highlighting problems, factors and how they become potentially influence the skills suggested in the past. Paper also provides some feasible solutions that are capable to deal with proposed challenges.

Social media is a powerful tool for people to share their thoughts and feelings. People post their personal feelings and thoughts on any topic, person, policy or product for marketing or social attraction. This text information can be broken down into facts and opinions. Reflect people’s feelings about government policies, products, national and international personalities, the constitution, incidents and events. Sentiment analysis is an area of study that analyzes people’s opinions, ratings, attitudes and emotions in general from the written language. The texts of the reviews are edited to give an accurate description of the author’s opinion on the subject. This paper proposes an analysis of the feelings of the writers about the subject and comments of the users of social networks in the context of Ethiopia with Facebook as a platform. In this paper, natural language processing (NLP) and machine learning (ML) based on a novel emotional AI sentiment analysis and opinion mining framework used to conduct sentiment analysis and opinion mining of posts and user comments on social media via a Facebook application.

The computer-aided diagnosis set the cornerstone for brain tumor detection. However, early-stage brain tumor detection is a very complex approach by the clinical process. The abnormal growth of cells in the brain forms the issue of the tumor. Therefore, the mortality rate of brain tumor patients is very high due to the late diagnosis of the tumor. On the other hand, the early-stage tumor saves millions of lives worldwide. This paper proposed a multi-sage support vector machine algorithm for the detection of tumors. The multi-stage support vector machine maximizes the area of margin and covers full features of MRI images. For the selection and optimization of features of tumors, apply the firefly algorithm. For the processing of features, extraction applies discrete wavelet transform. The proposed algorithm was tested with a reputed dataset such as BRATS. Furthermore, the proposed algorithm is compared with existing algorithms such as SVM and CNN. The proposed algorithm improves the detection rate by 2–3% instead of CNN and SVM algorithms. The evaluation of results was done under MATLAB software environments.

The aim of this work was to develop an AI-controlled fitness trainer to tend to each user’s needs. It includes an AI-based voice assistant that acts as a virtual fitness trainer to guide the user in performing a certain routine of exercises, which was implemented through the use of NLP to recognize the user’s voice for commands to activate the trainer and body pose recognition to monitor the user’s postures for the workouts in real-time. This work combined two different features and collectively helped the user to workout in a more efficient way. The trainer can currently perform these operations over a set of 10 workouts—7 for muscle workouts and 3 for cardio. Once the workout session is complete, a bar plot consisting of all the exercises performed during that session is constructed and stored on the user’s device. The study below goes into further detail on the major insinuations for future fitness coach design and assessment.

The aim of this work is to improve performance metrics for diagnosing coronary artery disease using machine learning (ML) classifiers. In this paper, voting ensemble learning, i.e., soft voting and hard voting was used by selecting five base ML classifiers: random forest, extra tree classifier, extreme gradient boosting, decision tree, and gradient boosting machine. Two datasets were used to test the ensemble model, and data pre-processing was carried out as a preliminary analysis. The prior features are also defined based on the voting approach. To prepare the dataset for modelling, the train-test split patterns are categorized to understand the dimensionality effect. From the computational evaluations, soft voting ensemble learning scored the highest score for both the datasets at every train-test split. The highest accuracy for dataset 1, and 2, is scored as 98.31%, and 90.09%, respectively, for the 90–10 splits.

Data mining is the method of analyzing and summarizing data in terms of useful knowledge, based on various perspectives. Classification is one of the approaches which is commonly used in medical data analysis. The goal here is to uncover new patterns in order to generate realistic results and knowledge about users. Currently, data mining approaches are applying to healthcare datasets to test effective strategies and this paper involves the application of various processes, steps, analyses and comparisons for the classifications of IVF datasets. A thorough performance analysis and comparative review of these approaches is conducted and can be used to choose the appropriate algorithm for future analysis for the given dataset.

In this paper, for web-based cloud computing services, we provide a novel fine-grained two-factor authentication access control mechanism. Our suggested 2FA access control system, which requires both a user secret key and a lightweight security device, employs an attribute-based access control mechanism in particular.

Organizations today are keen to adopt innovative marketing tools to ensure the gaps between consumer needs and preferences and business offerings which are bridged aptly to the extent that every notion of consumers is proactively understood and promptly served by the providers essential for optimizing business performance and profitability. This can be efficiently achieved by implementation and integration of advanced information technology tools and social media applications such as Twitter, Facebook, and Whatsapp that generate huge quantities of user data based on varied user needs, preferences, and interests. Gathering such user generated information could prove immensely beneficial for businesses today in precisely understanding consumer needs and proactively offering those needs in time which not only boosts performance of the enterprise but also ensures substantial cost cuttings on advertising, marketing, and promotions. From these perspectives, a unique approach to marketing that employs Naïve Bayesian algorithm would be investigated to examine its effectiveness and viability in marketing performance of businesses. The peculiarity of the chosen model to measure consumer views and opinions on products and services forms the basis of this study. Thus, the aim of this research is to conduct a sentimental analysis using social media posts using Naïve Bayesian algorithm in order to obtain realistic and predictable inferences based on consumer behavior.

India is mostly a farming country. Agriculture is the practice of using land to cultivate various types of crops. Approximately, 60% to 70% of India’s population is reliant on agriculture for a living. Farmers play a critical part in our country’s economic development; India ranks second in the world in farm outputs. Agriculture employed more than half of India’s workforce in 2019 and generated 17–18% of the country’s GDP. They are in the rural area. A rural area, sometimes known as the countryside, is a geographical location outside of towns and cities. In rural community, most of the people are not aware of the hygiene methods in their livelihood and some of the health issues. Even in some cases, farmers are not able to identify the leaf diseases in the crops they are growing; this could lead to less yield in the crops. By using technologies like IoT and ML, we can solve the problems faced by the rural community. So, we are creating a smart box, which is designed to facilitate communication between humans and machines. The machine has been programmed with the ability to recognize sentences and make decisions on its own in response to an inquiry. The response premise is to match the user’s input statement. The Raspberry Pi, camera, speaker, mic, and power supply were all utilized in this project. The Kannada language has been honed to the point where anytime a pharmaceutical name is displayed, it will indicate for what purpose it can be utilized. Our method also assists farmers by detecting leaf disease and prescribing appropriate treatments. It also monitors the level of hygiene in a given region and sends a notification to the individual who is concerned. Our proposed system includes an emergency dial for the ambulance, which assists persons in need of assistance.

In wireless sensor networks, the entire region is disbursed with a set of sensing devices (SDs), and then, multiple sets are formed. For controlling each set, a unique SD is selected for inter-set communication. The selection of special sensing device (SSD) in each region sensing device network (RSDN) is done by computing random value of probability for the case of LEACH method. Whenever end-to-end route is established, this SSD will keep on changing so that each SD has an opportunity to deliver data across regions. The end-to-end path formation is done using SDs, base controller (BC), and SSD. The disadvantage is that the path is longer, more hops, and more energy consumption. The E-LEACH improves the selection of SSD based on remaining energy levels of SDs. The route is formed using combination of SDs, primary, and secondary SD. This will overcome the delay and hop ratio disadvantage of LEACH but suffers from lower goodness network factor. The selection of SSD for a multi-criteria system is done by measuring remaining energy, distance with respect to reference position and mobile value. Comparison is done between LEACH, E-LEACH, and multi-criteria LEACH with respect various parameters, and it is proved that multi-criteria LEACH is the best.

Globally, the prostate is a second most frequent cancer in men and is the fifth leading cause of death. At early state, prostates are asymptomatic and do not require any care. However, if screened at early state, it can be removed immediately. The screening is done either with most widely used transrectal TRUS imaging or with magnetic imaging. The paper introduces a novel approach to demark the prostate boundary using spectral clustering approach with Gaussian similarity. The multimodal image database is obtained with the ethically approved collaboration. The algorithm is tested on 217 MRI images and 402 TRUS images. The competency of the approach was evaluated with respect to five quality metrics in both the modalities. The results showed that, the proposed approach is robust in demarking the prostate region with segmentation accuracy of 92.02 and $$93.78\%$$ 93.78 % for MRI and TRUS database. The highest average sensitivity and specificity of $$95.7\%$$ 95.7 % in segmenting MRI and TRUS images shows the robustness of the proposed SC approach in accurate delineation of prostate boundaries.

The use of deep neural networks is a recent research area in content-based image retrieval that has enhanced results on many datasets and outperformed handcrafted for fine-tuning of the network. The image feature is computed by combining different feature descriptors like shape with color and texture, so that it describes the image more effectively. In this paper, image retrieval on database images is done to get the top-k most similar database images using kNN on the image embeddings with cosine similarity as the distance metric. Two unsupervised methods are performed by generating image embeddings using a pre-trained network such as simple autoencoder, convolution autoencoder, and transfer learning such as VGG19, VGG16, and ResNET152. This is done by removing its last few layers and performing inference on our image vectors for the generation of flattened embeddings. No training is needed throughout this entire processing, only the loading of the pre-trained weights is needed. We train both a simple autoencoder and a convolutional autoencoder on our database images with the objective of minimizing reconstruction loss. After sufficient training, we extract the encoder part of the autoencoder and use it during inference to generate flattened embeddings. This work is using transfer learning and autoencoder-based retrieval system with classification. Experiments are carried out on flower and fruit datasets. Results are recorded using GPU on Google Colaboratory platform. Visualization is shown using “t-distributed stochastic neighbor embedding (t-SNE)”.

Attendance is an important part of the academic environment. The manual method of marking student attendance is time-consuming and also not accurate. So, the use of biometric attendance is a better alternative to the manual method. There are many biometric techniques that can be considered to design an automated system to mark attendance. Facial recognition is one such biometric technique that can be used. In this paper, we propose the implementation of facial recognition where the attendance is marked by recognizing the faces detected in the video feed from the classroom. We are in the midst of the once in a century crisis, ever since the COVID-19 pandemic broke out it has become imperative to accommodate certain behavioral changes in our day to day lives, one such major change which is essential to curb the spread of COVID-19 is to wear a face mask, and thus, the facial recognition-based attendance adds another advantage by recognizing the faces even though students would be wearing the masks. Another important measure that needs to be followed to contain the spread of COVID-19 is to ensure social distancing in all public spaces; hence, there is a need to ensure that social distancing norms are followed by the students. So, we propose implementation of a system to monitor the social distancing among the students. Further, we propose to implement a COVID-19 vaccination status monitoring system using which we can monitor the vaccination status of the individuals through the video feed from the classroom.

Achieving excellent performance in the fields of classification, detection, and recognition of objects in pictures and videos, Convolution Neural Networks (CNN) have been developed. Different Regularization Techniques have an impact on the performance of Convolution Neural Networks. By Applying Regularization Techniques such as Early halting, L2 Norm, Dropout, and data augmentation approach, we may improve the model performance significantly. The integration of all of these Regularization Techniques results in a drastic improvements in the performance of the neural network model. Model’s accuracy will improve up to 90%–95% and also reduces the over fitting problems, where results are predicting with respect to training model. This paper highlights the impact of various regularization techniques that boosts the learning convergence.

The corona virus pandemic has made severe changes to our lifestyle, thereby, indirectly affecting our physical and psychological well-being.: how we work, socialize. With the global pandemic still in the process of altering the definition of “normal life” across the planet, most industries are still scrambling to cope up. Electronic-commerce (e-commerce) is the buying and selling of services and products, or the transmitting of data or funds, over a network, primarily the web. Business to business (B2B), business to consumer (B2C), consumer to consumer or consumer-to-business are the different types of business transactions. The terms e commerce and e business are often interchangeable. The term e tail is used in addition to the reference in transaction processes for online shopping. E-Commerce platform with modern technologies which is MERN (MongoDB, Express, React, and Node.js) Stack. Agriculture specific functionality which incorporates selling agriculture related products. By this platform farmers will get exposure and farmers will receive their profits directly through the web site. E-Commerce platform includes PAN India support (including rural). Farmers will get a precise idea of what they are buying with the help of description provided by the admin. Ideal agro e commerce can make a remarkable benefit to “developing” countries. Make transactions risk-free and help track resource expenditure. Work as a powerful catalyst for economic development.

The performance of OFDM systems can be sharply deteriorated with effect of impulsive noise. It is proposed in this paper about the combined station impulse reaction approximation and impulsive noise mitigation process on the basis of compressed sensing concept. From these algorithms, we can treat channel impulsive responses and impulsive noise as joint sparse vector. Thereafter, framework of sparse-Bayesian learning is utilized for helping estimating the station impulse responses, impulsive noise and data signs. Here, data sign can be known as the unknown parameter. The Cramer Rao Lower Bound’s elaborated as the bench-mark. From all used impulse noise-mitigation techniques, proposed procedure in the paper can be used with the sub-pilot by not using past info of the station and impulsive noise. The model outcomes conclude at the process projected in the paper improves the efficiency in the station approximation and enhances performance of bit error rate.

Many adolescent people are fighting against to corruption, black money and are trying to make a difference. People interested to tackle these mentioned problems and need guidance can refer this paper. The main intention behind this paper is to bring a direct connection between the people and the government, thereby reducing the involvement of many intermediary people and thus not making them involved to indulge in activities such as making black money and also as people can directly reach out to government to talk about the problems faced by them, it takes very less time for government to take initiative against the problem. This paper mentions about how to resolve the problem of black money and ongoing issues related to state governments of India. In the future, the same idea can be applied to solve related problems.

The transfer of the data from one Sensing Intelligent Device (SID) to the other sensing intelligent devices (SIDs). The transmission of data happens from one area to a different area in a Sensing Intelligent Network (SIN). The data packets receiver is usually the destination Control Center. Various applications of SIN are disaster monitoring, war plane detection and sending the environmental hazards information along with various other applications. Due to the constraints which are involved in SIN like low battery, low transmission range it is important to avoid duplication of data as well as collect the data from the SIDs in an intelligent fashion. Each SIN has huge number of SIDs and hence they transmit data which is similar in nature and hence this causes the SIDs to choke and become irrelevant SIDs over a period. The number of such chocking instances can be reduced in the SIN with the help of data aggregation, intelligent selection of centralized SID for a specific SIN area with the help of machine learning techniques. This paper lists out various data aggregation techniques which are available in the literature and how they can be used to improve the relevant ratio value for SIN.

In many countries, health examination is a part of healthcare system. Distinguishing the members at threat is essential for precautioning and for taking preventive measures. The main problem of understanding a classification method for risk indication is in the unknown data that forms a greater part of gathered dataset. Especially, the unlabelled data specifies the members whose health states can change greatly. This paper is about an algorithm called semi-supervised heterogeneous graph (SHG) on health for disease identification.

In this paper, three popular model predictive control (MPC) techniques, namely predictive current control (PCC), predictive torque control (PTC), and predictive flux control (PFC), applied to induction motor (IM) drive are presented. These techniques directly use the inherent discrete nature of the power converter and evaluate control parameters for the finite switching states. These control techniques for IM can offer fast dynamic response, improved steady-state operation, and retain the decoupled nature. In case of PTC, weighting factor is used in cost function, whereas in other two control techniques, weighting factor is only required when additional control parameters are added. These control techniques are developed in MATLAB/SIMULINK, and results are presented for various operating conditions.

Transliteration techniques are used to preserve the phonetic information associated with a word written in one language called the source language and write the word in another language called the target language with the pronunciation constrained of the target language. In this paper, a comparative study of English and Telugu language in terms of transliteration problem is presented, and a simple phrase-based machine translation (PBMT) system is trained using Moses and Giza++ for English to Telugu transliteration of general English words used in the English language. To evaluate the results, a modified edit distance-based method is used.

Given a connected, edge-weighted and undirected graph, the degree-constrained minimum spanning tree (dc-MST) problem seeks to find a minimum cost spanning tree (T) subject to every vertex in T has at most d degree, where d is a specified integer. The dc-MST problem is $$\mathcal{N}\mathcal{P}$$ N P -hard for $$d \ge 2$$ d ≥ 2 and finds several practical applications. This paper presents a hybrid approach (hABC) combining an artificial bee colony algorithm with two local search strategies for the dc-MST problem. The proposed hABC has been tested on a set of available benchmark instances and experimental results show its superiority overall against state-of-the-art approaches.

This paper gives a concise review on mathematical modeling of brushless direct current (BLDC) motor, the challenges on BLDC motor, torque ripple reduction techniques, condition for minimum commutation torque ripple, and different control strategies for torque ripple reduction in sensor and sensorless BLDC motor. This review also discusses the comparison between sensor and sensorless BLDC drives, merits of sensorless BLDC drive, and outcomes of different torque ripple reduction techniques. It also gives an idea to select better control strategy for specific BLDC drive applications. The best-suited control strategies have the advantages such as high efficiency, minimum torque ripple over the entire speed range, less switching loss, simple modulation scheme, improvement in DC voltage utilization, smooth and noiseless operation, and regulation in the speed of BLDC motor. The MATLAB simulation tool is used to verify the results of few topologies.

Two key fields that are extensively employed for data security are cryptography and steganography. With the help of these technologies, data security is offered in the banking system. The proposed method uses a hash function to virtualize bank cheque transactions. The cryptographic hash function algorithm is used to encrypt information. This paper uses steganography to develop an efficient algorithm for embedding data in images, which provides a better security pattern for transferring messages over a network. To achieve a secure transfer, the sender and receiver’s authentication details are masked. The projected method improves the integrity and confidentiality of the system.

The swarm intelligence algorithms are established as comprehensive approaches for resolving the complicated problems in optimization through the simulation of “behaviors of the biological swarms.” Recently, data mining areas seek more attention that requires analysis of huge data and quick management. Most of the traditional techniques have not provided effective performance, and they are only applicable for differentiable and continuous functions. Further, the group of population-based techniques is proved to be effective in certain researches, where the swarm intelligence algorithms provide the significant potential for relevant data mining tasks. Hence, this paper demonstrates the biological inspiration and some of the swarm intelligence concepts, in which the main focus is given to “Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO) algorithms.” The primary data mining strategies are also described along with certain existing and present works based on swarm intelligence approaches.

Video-based face recognition is likely exigent for the reason that a large amount of data has to be processed and it involves intricate computations. This paper provides a novel video-based face recognition system, designed using the convolution neural network (CNN) algorithm. The CNN is trained with the database generated from a video stream to recognize the drug dealers. If a drug dealer is detected, the system alerts the school or college management. The proposed drug distributor detection system is implemented using a 64-bit raspberry Pi 3 microcontroller and a Pi camera module that captures the video stream in real-time. If the detected face is a drug dealer, the system alerts the school or college management through SMS. The experimental results show that the CNN is capable to classify authorized students and drug dealers correctly and detect abnormal activities without using huge amount of training data. A top recognition accuracy of 99.58% was acquired for CNN.

Malaria disease is rapidly spreading and very dangerous disease. It is caused due to bite of mosquito. Numbers of peoples all over the world are infected by this disease. Traditionally, the microscopic method could be considered as a best method to detect the malaria from cell images, but it requires more time and it need skilled and expert person in that field so, this research is focused on developing CAD system to diagnose the malaria patient via deep learning model which gives automated accurate and faster results than traditional system. Objective of this paper is also too aware the people about symptoms, prevention techniques and treatment usually used to cure this disease. Early prevention and treatment may save number of deaths and infections. This research uses deep learning-based transfer learning technique to make CAD system with the help of flask app. Malaria detection is Web app that can be used anywhere any time and this method gives accuracy of 99.6% with fine-tuning approach of ResNet50 model.

Recent wireless technology boons smart solutions and offers high networking pliability. Through wireless sensor networks, users can access their information remotely at any time and anywhere. In vehicular ad-hoc networks (VANETs), the nodes that communicate among themselves to share the information have limited energy and resources. Moreover, as the nodes are highly dynamic in nature, i.e., able to leave or join the network arbitrarily, which increases network power complexity and design challenges in this type of network. Consequently, several vehicular ad-hoc routing protocols have been evolved to conserve energy, improve network performance, and address the encountered design challenges. The literature presents an ample variety of techniques and models that have been proposed for efficient, power-aware optimization in VANETs. In this paper, we propose a new taxonomy that categorizes the most common VANET routing perspectives published in the literature. The paper presents a comparative exploration of power optimization approaches to better understand the current research directions in VANET power optimization in routing.

The Raspberry Pi-based reader for the blind is based on the Raspberry Pi processing board. The primary goal of this initiative is for the blind person to be able to understand the context with in paper without the assistance of others. The Python script may be used to create this. Previously, visually handicapped persons used only to understand the context through understanding braille. Braille would be a writing method for blind that employs slightly elevated characters. Under this procedure, the blind person must learn Braille or rely on the assistance of another person for reading or understanding the context. To prevent such issues, consider the “Raspberry Pi-enabled smart reader for blind”. This is developed by incorporating devices such as a speaker, camera that serve as a gateway between both the user and the system. Throughout this work, optical character recognition (OCR) would be used to detect characters, which would then be read aloud by the system via a speaker. Whenever a piece of paper is put in front of the camera, an image of the content in the paper is taken. Following that the text would be transformed into voice and read out by a speaker.

In teaching–learning process, student attentiveness throughout the session greatly influences the quality of the acquired knowledge. In normal class room circumstances, teacher can identify attentiveness of the students by observing their facial expressions and body language. It imposes lot of burden on the teacher who wish to teach in effective manner. Often this method may not give accurate results, as it is very difficult to measure level of attentiveness of all the students by mere observing their faces. This paper discusses design and development of a model which evaluate mind status of each individual student based upon attention parameter. In present work, brain sensors are used to capture the EEG signals from the students using non-invasive method and Bluetooth Hub. Software is developed, which can give instantaneous report of all the students simultaneously after processing the brain signals received from the sensors. Test results of some trials are discussed.

In most of the medical applications, the accuracy of detecting and diagnosing the disease in a proper procedure is always a challenging issue. One of the most searched research works is brain tumor detection with most effective way; here, deep learning-based algorithms yield better outcomes. Brain tumor segmentation algorithms have been more investigated in the recent times with varied algorithms for accurate segmentation and detection. In this work, brain tumor detection from the MR scan images is mainly based on convolution neural network (CNN) amalgamated with multi-SVM classifier with various preprocessing and intermediate steps involved to bring out optimal results. In preprocessing stage, image filtering and intensity normalization of input images are carried out. At later stages, CNN along with multi-SVM classifier is utilized for training, testing along with classification process. Finally, the classification tends to display the presence of brain tumor or not. The implementation is processed by means of MATLAB computing language with necessary prerequisites.

The images that are captured underwater contain various artifacts and also degraded due to the inherent properties of the location parameters such as dispersion and refraction. In general, most of the fusion algorithms in image processing applications are mostly based on fusion of two images, but in this work, it is only one image approach, where the same image is converted into varied forms of two images based on color balance and white balance procedures. The proposed method indicates that images are enhanced by means of global contrast and also preserving the edges of the images with little modifications. Basic filtering operations are also utilized in this work to enhance the images in order to retain the integrity of the images to identify the underwater objects vividly.

Pulse-coupled neural networks are a subpart of deep learning (DL) methodologies, which has vast number of applications. One of the preferred applications is multimodal medical image fusion, where different modality images such as MR scan images, CT scan images, and PET scan images are needed to be fused in one of the combinations to provide promising results to diagnose the abnormalities that were unable to detect in the individual images. When accomplishment a precise diagnosis or anatomy of the body, one type of medical imaging modality may not be adequate, imposing the fusion of images from several medical imaging modalities in order to deliver a final result in the immense popular of medical applications. In this work, MRI and PET scan images are fused by means of PCNN and shearlet transformation to yield better results.

This research paper investigates and experiments with millimeter-wave-based orthogonal frequency division multiplexing (OFDM) uplink transmission. The spectral efficiency of the system is examined by taking the signal-to-noise ratio (SNR) into account in the channel model. The research and simulation findings reveal that the number of antennas has an effect on spectral efficiency values and SNR values. The spectral efficiency values were examined under three conditions: interference, no interference, and approximation.

In regards of switching frequency and energy efficiency, QCA technology would be thought to be a viable solution for electronic circuit. A MUX may be considered a viable option for constructing QCA circuits. Throughout this study, a distinct topology of energy-efficient QCA-based 2 × 1 MUX is suggested. This MUX exceeds the best available design in regards to energy usage. Furthermore, same or superior performance characteristics are attained compared to the present solutions. A MUX design might be used to replace majority-based designs in QCA as a significant basic fundamental piece. The recommended MUXes have an excellent scaling function and may be used in power-complex QCA circuits.

Quantum dot cellular automata (QCA) is regarded among the most key techniques for replacing existing CMOS technology. In contrast to standard transistor technology, the computing is built on a novel model deals with the interaction of neighbouring QCA cells. It has many benefits, including a high operating frequency (THz), an extremely high density, as well as low energy consumption. This study presents and constructs a novel XOR/XNOR logic gate using QCA nanotechnology. Performance is examined and studied to illustrate the functions and flexibility of the suggested QCA-based XOR/XNOR design. In contrast to various current QCA-based XOR designs, the suggested XOR/XNOR logic gate outperforms them in terms of area and energy consumption. Furthermore, QCA is used to construct several efficient circuitry based on the suggested XOR/XNOR gate.

QCA is a transistor-free method of realizing nanoscale circuit architectures. When compared to the commonly utilized CMOS technology, QCA circuits perform faster, more dense, and consume less energy. In this study, a new digital comparator structure based on QCA nanotechnology is suggested. The digital comparator, that contains 2 binary integers, is a fundamental and crucial module of the CPU. As compared to previous designs, the suggested digital comparator is optimum, single-layered, and contains less QCA cells. The suggested digital comparator has been evaluated to current digital comparators for several performance parameters. The suggested coplanar comparator has been built using the smallest QCA cells possible, resulting in a smaller total cell area and total coverage area. As a result of the low dissipation of energy for the suggested design, the suggested digital comparator becomes extremely energy efficient.

The significant advancement in medical imaging methodology is a multimodal medical image fusion method. Its primary goal is to provide a complete overview of medical image fusion methods, including theoretical foundations and recent breakthroughs. The primary goal is to gather useful information by merging several images obtained from different sources into a single image appropriate for better diagnosis. Magnetic Resonance Imaging (MRI) aids professionals’ decision making during the aided diagnostic pipeline in medical techniques such as Computed Tomography (CT). The fused image, on the other hand, may aid in the performance of other tasks such as classification, detection, and segmentation. The suggested technique eliminates distortion from the source image in the first step, followed by image enhancement, weight extraction, weight map computation and refining, pyramid decomposition, and output fused image. Matlab programming tool is utilized for this procedure, and the authors can customize the parameter settings. In this experiment, we focus on qualitative rather than quantitative analyses. The method’s conclusion is that current multimodal medical image fusion study findings are more relevant and might be used to successfully diagnose patients.

An analytical methodology for channel estimation and data decoding in huge multiple input multiple output uplink systems using 1-bit analog-to-digital converters is given (ADCs). Various approaches have been developed, but the quadrative amplitude modulation (QAM) method is the most commonly employed. The receiver had to identify the “amplitude and phase” of each incoming symbol in order to decode QAM. When the signal intensity is greater than the noise strength, the two ends may choose the constellation procedure. The receiver in two-way communication has an equalizer and must untwist the incoming symbols back to their intended shape in order to decipher them accurately. Existing work does not provide a clear expression of the mean square error (MSE) of channel estimation, making it impossible to examine its system performance in relation to various factors. Because it necessitates an extremely linear amplifier. In this work, MSE, symbol error rate (SER), and variance values are computed and provided their analysis to estimate the system performance.

The various bad climatic whether conditions occur in our daily life are snow, sandstorm, haze, fog etc., which will effects the natural scene and the visibility and these will be restore by image dehazing. In this work, we compute the relativity of Gaussian for red, green, and black color spaces to produce guidance image and improved dark channel prior is used to improve the transmission depth map construction in our work. There are different dehazing methods used to restore the visual perception of the image, increase the natural view. It is a highlighted research area because of its real-time applications in surveillance systems, driver assistance system especially for the people who resides in hilly areas where mist and haze is prominent. The proposed dehazing method is based on improved dark channel prior and vanherk model which is concentrated on detailed enhancement of the transmission map and hence reconstructing the hazeless image. The different proposed dehazing methods results in both qualitative analysis and quantitative analysis. Quantitative analysis has more existing methods like structural similarity index and peak signal to noise ratio these are used for the evaluate when ground truth images are present and blind reference less image spatial quality evaluator and naturalness image quality evaluator are used to evaluate when ground truth images are not present. These existing methods give better results without any distorted images. This work provides good performance for the task of obtaining hazeless image both qualitative and quantitative manner. Experimental results show that the proposed method estimates haze more accurately, the reconstruct the images are more realistic and detailed information is restored.

Infrared and visible image fusion expects to obtain images that highlight the thermal radiation information from infrared images and texture details from visible images. In this work, a novel fusion of Bayesian model is established for visible and infrared images. In our model, the image fusion task is cast into a regression problem. To measure the uncertainty in a better manner, the model is formulated in a hierarchical Bayesian manner. Aiming at making the fused image satisfy the human visual system, the model incorporates the total variation (TV) penalty. In the test phase, the two decomposition feature maps of base and details are merged, respectively, by the fusion layer, and then, the decoder outputs the fusion image qualitative and quantitative results demonstrate that our method can generate fusion images containing highlighted targets and abundant detail texture information with strong reproducibility and meanwhile superior than the state-of-the-art. In this work, MATLAB R2017b version tool is utilized for simulation purpose.

Optical coherence tomography (OCT) imaging is a precise and significant approach in retinal diagnostics at the layer level. The diseased effect in the retina poses a barrier to a computational segmentation approach at the boundary layer level for defect evaluation and diagnosis. The noise in the computing approach misguides the layer segmentation and border edging operation. In these requirements, a novel segmentation algorithm based on a denoising technique is required. The many layers of OCT under the macula area of the eye are to be highlighted in this work. The proposed contour model is used to determine the nerve fiber layer (NFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer nuclear layer (ONL), inner segment (IS), outer segment (OS), and retinal pigment epithelium (RPE). As preprocessing approaches, the median filter and histogram equalization are utilized, and the canny edge detector, coupled with watershed thresholding, provides post-processing so that layers can be easily identified. For simulation purposes, the MATLAB R2017b version tool is used in this work.

Because of its relatively high and license-free bandwidth, great security, and low cost, visible light communication (VLC) is a strong competitor for vehicle VLC (V2LC). Its durability, high spectral efficiency, and ability to deal with inter-symbol interference make it an excellent choice for autonomous and connected vehicle applications that demand high data rate transmission (ISI). VLC systems can benefit from high-rate communication by reducing pilot overhead in standard pilot-based CE methods. However, this comes at the expense of greater complexity and decreased CE accuracy, which are both undesirable outcomes. Techniques for blind medium estimation (CE) for OFDM systems are discussed in the radio-frequency (RF) literature. There has been no work done in the VLC literature, however, on blind CE for OFDM systems. In order to improve CE accuracy for V2LC, we propose a new blind CE technique based on real-time data-based medium characteristics that is applicable to V2LC. The normalized medium frequency response (CFR) of V2LC broadcasts has been found to be unaffected by inter-vehicle distance, relative transmitter/receiver (Z) zenith angles, or even ambient light, according to real-world data collected from real cars to vehicles. As an alternative to estimating the normalization factor at each subcarrier individually, this medium characteristic is used to estimate the normalization factor for calculating the normalization factor in the blind CE. After running a large number of simulations at a variety of vehicle speeds, it was discovered that the suggested strategy beats pilot-based CE systems in terms of average throughput and bit error rate for all modulation schemes (BER). Moreover, for the realistic vehicle mobility scenario chosen from the Simulation of Urban Mobility, the real-time performance of the proposed blind CE is proven to be very close to the maximum throughput of each modulation scheme at high signal-to-noise ratio (SNR) levels (SUMO).

Diabetic retinopathy (DR) is defined using progressive identification of the retina which appears in various varieties of retinal disease such as microaneurysms, hemorrhages, exudates, and so on. The detection of these retinal diseases is critical for the early diagnosis of DR. In order to detect these retinal diseases at an early stage, we proposed in this project to develop an efficient retinal layer segmentation and clustering approach based on the robust Deep Learning Regression technique for early detection and diagnosis of retinal diseases such as glaucoma, macular degeneration (age-related), diabetic retinopathy, and cognitive degeneration disease such as dementia by observing changes in the retinal layer thickness. For the upcoming method involvement, we are undergoing regression techniques. We are going to pre-process the input images, and by using advanced filtering techniques, we are going to remove the artifacts which might be added to the input images during the scanning process. We cluster and segment the images with edge detection. Then, we try to feature extract the significant attributes of layers and classify what type of diseases got affected to the eye. For effective efficient output, we are going for the deep learning processes where we train the features for betterment purpose. The proposed methodology will be applied to both OCT and FUNDUS images, and the performance of retinal disease detection and classification will be evaluated for both classes of images.

IC technology advances daily in order to improve device efficiency and density of small devices. For the last 4 decades, standard CMOS technology has been a critical component of digital computing. However, scaling CMOS systems has been a struggle over the previous several years. QCA has indeed been recognized as a unique nanoelectronic technology. To provide a novel idea of integrated circuit design in an effective and optimal way, an effective design for BCD adders in QCA technology has been given here, utilizing an area optimized QCA complete adder. In compared to previous BCD adder design, the suggested BCD adder circuit optimizes QCA design parameters including such layout area, as well as number of QCA cells.