Skip to main content

2024 | Buch

Computational Sciences and Sustainable Technologies

First International Conference, ICCSST 2023, Bangalore, India, May 8–9, 2023, Revised Selected Papers

herausgegeben von: Sagaya Aurelia, Chandra J., Ashok Immanuel, Joseph Mani, Vijaya Padmanabha

Verlag: Springer Nature Switzerland

Buchreihe : Communications in Computer and Information Science


Über dieses Buch

This book constitutes the revised selected papers of the First International Conference, ICCSST 2023, held in Bangalore, India, during May 8–9, 2023.

The 39 full papers included in this volume were carefully reviewed and selected from 200 submissions. They focus on artificial intelligence, blockchain technology, cloud computing, cyber security, data science, e-commerce, computer architecture, image and video processing, pandemic preparedness and digital technology, pattern recognition and classification.


Performance Evaluation of Metaheuristics-Tuned Deep Neural Networks for HealthCare 4.0

The emergence of novel technologies that power advanced networking, coupled with decreasing sizes and lower power demands of chips has given birth to the internet of things. This emerging technology has resulted in a revolution across many fields. A notably interesting application is healthcare where this combination has resulted in Healthcare 4.0. This has enabled better patient monitoring and resulted in more acquired patient data. Novel techniques are needed, capable of evaluating the gathered information and potentially aiding doctors in providing better outcomes. Artificial intelligence provides a promising solution. Methods such as deep neural networks (DDNs) have been used to address similarly difficult tasks with favorable results. However, like many modern algorithms DNNs present a set of control values that require tuning to ensure proper functioning. A popular approach for selecting optimal values is the use of metaheuristic algorithms. This work proposes a novel metaheuristic based on the sine cosine algorithm, that builds on the excellent performance of the original. The introduced approach is then tasked with tuning hyperparameter values of a DNN handling medical diagnostics. This novel approach has been compared to several state-of-the-art algorithms and attained excellent performance applied to three datasets consisting of real-world medical data.

Luka Jovanovic, Sanja Golubovic, Nebojsa Bacanin, Goran Kunjadic, Milos Antonijevic, Miodrag Zivkovic
Early Prediction of At-Risk Students in Higher Education Institutions Using Adaptive Dwarf Mongoose Optimization Enabled Deep Learning

The biggest problem with online learning nowadays is that students aren’t motivated to finish their coursework and other assignments. As a result, their performance suffers, which raises the dropout rate, necessitating the need for proactive measures to manage the dropout. Predictions of student performance assist in selecting the best programmers and designing efficient study schedules that are suited to their needs. Additionally, it aids in the development of observation and support tactics for students who require assistance in order to finish the course work by teachers and educational institutions. This paper proposed an efficient method using Adaptive Dwarf Mongoose Optimization (ADMOA)-based Deep Neuro Fuzzy Network (DNFN) for prediction of at-risk students in higher education institutions. Here, DNFN is working to forecast at-risk kids and prediction is carried out based on the most pertinent features collected utilizing the created ADMOA algorithm. Additionally, the effectiveness of the proposed ADMOA_DNFN is examined in light of a number of characteristics, including Root MSE (RMSE), Mean Square Error (MSE), Mean Absolute Error (MAE) and Mean Absoulte Percentage Error (MAPE), it attains best values of 0.049, 0.045, 0.212, and 0.022 respectively.

P. Vijaya, Rajeev Rajendran, Basant Kumar, Joseph Mani
Decomposition Aided Bidirectional Long-Short-Term Memory Optimized by Hybrid Metaheuristic Applied for Wind Power Forecasting

Increasing global energy demands and environmental concerns have in recent times lead to a shift in energy production towards green and renewable sources. While renewable energy has many advantages, it also highlights certain challenges in storage and reliability. Since many renewable sources heavily rely on weather forecasting the amount of produced energy with a degree of accuracy becomes crucial. Energy production reliant on wind farms requires accurate forecasts in order to make the most of the generated electricity. Artificial intelligence (AI) has previously been used to make tackle many complex tasks. By formulating wind-farm energy production as a time-series forecasting task novel AI techniques may be applied to address this challenge. This work explores the potential of bidirectional long-short-term (BiLSTM) neural networks for wind power production time-series forecasting. Due to the many complexities affecting wind power production data, a signal decomposition technique, variational mode decomposition (VMD), is applied to help BiLSTM networks accommodate data. Furthermore, to optimize the performance of the network an improved version of the reptile search algorithm, which builds on the admirable capabilities of the original, is introduced to optimize hyperparameter selection. The introduced method has been compared to several state-of-the-art technique forecasting wind energy production on real-world data and has demonstrated great potential, outperforming competing approaches.

Luka Jovanovic, Katarina Kumpf, Nebojsa Bacanin, Milos Antonijevic, Joseph Mani, Hothefa Shaker, Miodrag Zivkovic
Interpretable Drug Resistance Prediction for Patients on Anti-Retroviral Therapies (ART)

The challenge of eliminating HIV transmission is a critical and complex under taking, particularly in Africa, where countries like Uganda are grappling with a staggering 1.6 million people living with the disease. The virus’s fast pace of mutation is one of the main challenges in this battle, which often leads to the development of drug resistance and makes it difficult to provide effective treatment through AntiRetroviral Therapies (ART). By leveraging the latest innovations in Smart Technologies and Systems, such as Machine Learning, Artificial Intelligence, and Deep Learning, we can create novel approaches to tackle this issue. We presented a model that predicts which HIV patients are likely to develop drug resistance using viral load laboratory test data and machine learning algorithms. On the remaining 30% of the data, we tested our algorithms after painstakingly training and validating them on the previous 70%. Our findings were remarkable: the Decision Tree algorithm outperformed four other comparative algorithms with an f1 scoring mean of 0.9949, greatly improving our ability to identify drug resistance in HIV patients. Our research highlights the potential of combining data from viral load tests with machine learning techniques to identify patients who are likely to develop treatment resistance. These findings are a significant step forward in our ongoing fight against HIV, and we are confident that they will pave the way for new, innovative solutions to address this global health crisis.

Jacob Muhire, Ssenoga Badru, Joyce Nakatumba-Nabende, Ggaliwango Marvin
Development of a Blockchain-Based Vehicle History System

A complete vehicle history can be obtained by using the recording, storing, retrieving, and tracking of pertinent information re garding a vehicle throughout its entire lifecycle, beginning with its initial sale by the manufacturer/dealer to its ultimate disposal as scrap. All par ties in the Registration Chain, including the vehicle supplier, owner, various government agencies, and the general public, can have confidence, trust, and authentication of records kept by the agency by incorporating a tamper-proof approach like blockchain to capture this crucial information in a distributed ledger register. A decentralized, secure, digital ledger that documents such transactions, avoiding or minimizing the limitations of conventional methods, can be advantageous to all parties involved in the Vehicle Registry. Traditional systems also lack the ability to con firm whether any data has been altered or falsified, a problem that can be addressed with blockchain’s trusted data. Decentralizing the Registrar process, enhancing data accessibility, and boosting security are all possible with the implementation of a vehicle registration ledger system based on blockchain technology. This paper explores the development of blockchain-based vehicle history system in Oman.

Mani Joseph, Hothefa Shaker, Nadheera Al Hosni
Social Distancing and Face Mask Detection Using YOLO Object Detection Algorithm

Due to the COVID-19 pandemic, there has been a huge impact worldwide. The transmission of COVID-19 can be prevented using preventive measures like social distancing and face masks. These measures could slow the spreading and prevent newer ones from occurring. Social distancing can be followed even by those with weaker immune systems or certain medical conditions. With the new normal into play, maintaining distance in social and wearing masks are likely to be followed for the next two years. This paper studies about maintaining distance in social and detection of masks using deep learning techniques. Several object detection models are used for detecting social distance. The inputs used are in the form of images and videos. With this system, the violations can be detected which will reduce the number of cases. In conclusion, the proposed system will be very efficient and can also be used to introduce newer preventive measures.

Riddhiman Raguraman, T. S. Gautham Rajan, P. Subbulakshmi, L. K. Pavithra, Srimadhaven Thirumurthy
Review on Colon Cancer Prevention Techniques and Polyp Classification

Colorectal Cancer is the second largest type of life-threatening disease in humanity for a long time. The major causes of CRC are enlisted. Colonoscopy followed by the polypectomy during the procedure is the known method adopted for survival. Identifying the precancerous polyp is the challenging task as well as for disease-free survival. For this alone a few deep learning as well as profound learning strategies are recommended for the analysis, classification, and identification of the object that is generally in the form of a polyp. The detection of the presence of polyp for the given set of images materializes the success of the proposed process model that caters to the neural networks algorithm.

T. J. Jobin, P. C. Sherimon, Vinu Sherimon
Security Testing of Android Applications Using Drozer

Android applications are extensively utilized however, many of them include security flaws and malware that serve as entry points for hackers. Drozer is used in the current study to evaluate the security of four vulnerable Android applications namely AndroGoat, Diva, Insecure Bank v2 & BeetleBug. The information security knowledge of undergraduate students in Oman was evaluated, in the current study, using an online questionnaire. Where, A virtual emulation environment was employed to run a penetration test and examine the attack surfaces of four vulnerable Android applications. Participants in the study showed high level of security awareness when it comes to managing the applications and their permissions as well as posting personal information. The studied packages included a range of exporting components (Activities, content providers, services and broadcast receivers) that are not particularly covered by constraints, making them susceptible to hacking and data exploitation and potentially posing a security risk. Reducing attack surfaces in apps requires taking measures like defining permissions, executing authentication procedures during intents transition, securing databases, and cleaning data after usage. Using four unsecure Android applications, this study categorized Android vulnerabilities based on the OWASP mobile 2016 risks. This research is recognized as an adjunct model that security experts, researchers, and students may use to identify vulnerabilities and assure application security.

Kamla AL-Aufi, Basant Kumar
Contemporary Global Trends in Small Project Management Practices and Their Impact on Oman

Globalization and internationalization bring change in the business domain. Competition expands its roots in international dimensions and corporations are emerging. Organizations are now adopting new trends and accepting changes to win the market position. Due to global trends, the management of projects is now using different methods, tools, and practices to make their project successful. This study is exploring the impact of global trends on the management practices of small projects. A background study was done to analyze the past situation, moves in the traditional practices, and make a link with current global trends. The interview was conducted with the leaders of small projects to get an in-depth view of global trends and their impact on management practices. Highlighted global trends are digitalization, sustainability, diversity, people analytics, research and development, shared knowledge, and innovation. The second stage of research checks the awareness level of global trends and practices. Data was collected from the people working on small projects. These people are well aware of global trends and have a ready culture in them. Proper planning, supervision, and resources are needed to flourish small projects as these projects are a good contributor to Oman’s economy and consider an important component of the business world. This study opens a new way for the research to explore industry-wise small project management practices and analyze their pattern of meeting global demand.

Safiya Al Salmi, P. Vijaya
Early Prediction of Sepsis Using Machine Learning Algorithms: A Review

With a high rate of morbidity as well as mortality, sepsis is a major worldwide health concern. The condition is complex, making diagnosis difficult, and mortality is still high, especially in intensive care units (ICUs), despite treatments. In fact, the most common reason for death in ICUs is sepsis. As sepsis progresses, fatality rates rise, making prompt diagnosis essential. Rapid sepsis detection is essential for bettering patient outcomes. To help with early sepsis prediction, machine learning models have been created using electronic health records (EHRs) to address this issue. However, there is currently little use of these models in ICU clinical practice or research. The objective is to use machine learning to provide early sepsis prediction and identification utilizing cutting-edge algorithms. For adult patients in particular, the creation of very precise algorithms for early sepsis prediction is crucial. Advanced analytical and machine learning approaches have the potential to predict patient outcomes and improve the automation of clinical decision-making systems when used in Electronic Health Records. The machine learning methods created for early sepsis prediction are briefly described in this paper.

N. Shanthi, A. Aadhishri, R. C. Suganthe, Xiao-Zhi Gao
Solve My Problem-Grievance Redressal System

This study aimed to provide a platform that would be an interactive platform for the local common people to lodge their grievances and complaints. The grievances (such as potholes on the roads, garbage on the streets, sewage leakage, etc.) are raised exponentially with the rapid growth of the population. These grievances and complaints have severely. Government officials and developers worldwide strive for measures to control problems and complaints encountered by people in day-to-day life. The traditional grievance procedure or system is time-consuming, and people in remote areas frequently lack assistance. In existing grievances, redressal application complaints are lodged after registration through many forms; an alternative to that would be manually filing the complaint by visiting the official departmental office. To provide a recognized platform that enables proper communication between citizens and locally elected members, we proposed the application “Solve My Problem,” which accepts complaints in a variety of different formats, such as text, images, or video based. The app is used to submit grievances and complaints, which are then displayed to the user’s locally elected representative. In comparison to the existing grievance redressal system, the app doesn’t need long procedures of form filing. It requires just a click of a photo of the issue. The proposed application then uses an ML model that has been trained on image datasets of common problems like pothole, sewers, garbage etc. for classification and extracts location content of the image to trace the location of the issue. The grievances are filtered based on the number of grievances in each area, which are then passed on to higher authorities. Using data visualization techniques that can be viewed by the authorities as well as by the citizen who filed the complaint, there will be a transparent process, and complaints can be tracked. Using the proposed application, the aim is to fast-track the process of filing grievances, connecting the citizen and the government and saving lives by fixing small problems.

Sridhar Iyer, Deshna Gandhi, Devraj Mishra, Rudra Trivedi, Tahera Ansari
Finite Automata Application in Monitoring the Digital Scoreboard of a Cricket Game

Cricket is one of the most viewed sports in the world and tracking of this game is very important throughout the gameplay. The application of automata in sports is least explored and this paper proposes the use of automata theory in cricket. The paper sheds light upon the potential application of finite automata concept in cricket which proposes proper tracking flow of the game. The major focus of the paper is shown upon ball-tracking and on-strike batsman tracking with each tracking system being explained in detail using the state diagrams. With the help of Deterministic Finite Automata (DFA), it is possible to achieve the flow of track and thus, helps in efficient tracking of the sport. It also minimizes the rate of committing the tracking errors during the gameplay.

K. Kaushik, L. K. Pavithra, P. Subbulakshmi
Diabetes Prediction Using Machine Learning: A Detailed Insight

Diabetes often referred to as Diabetes mellitus is a general, continuing and deadly syndrome occurring all over the world. It is characterized by hyperglycemia which occurs due to abnormal insulin secretion which results in an irregular rise of glucose level in the blood. It is affecting numerous people all over the world. Diabetes remained untreated over a long period of time, may include complications like premature heart disease and stroke, blindness, limb amputations and kidney failure, making early detection of diabetes mellitus important. Now a days in healthcare, machine learning is used to draw insights from large medical data sets to improve the quality of patient care, improve patient outcomes, enhance operational efficiency and accelerate medical research. In this paper, we have applied different ML algorithms like Logistic Regression, Gaussian Naive Bayes, K-nearest neighbors, Support Vector Machine, Decision Tree, Random Forest, Gradient Boost, AdaBoost and Multi Layered Perceptron using Artificial Neural Network on reduced PIMA Indian Diabetes dataset and provided a detailed performance comparison of the algorithms. From this article readers are expected to gain a detailed insight of different symptoms of diabetes along with their applicability in different ML algorithms for diabetes onset prediction.

Gour Sundar Mitra Thakur, Subhayu Dutta, Bratajit Das
Empirical Analysis of Resource Scheduling Algorithms in Cloud Simulated Environment

The cloud environment is a collection of resources providing multiple services to the end-users. Users submit tasks to this cloud computing environ ment for computation purposes. Using statically fixed resource scheduling algo rithms, the cloud accepts these tasks for computations on its Virtual Machines (VM). Resource scheduling is considered a complex job, and computing these challenging tasks without external intelligence becomes a challenge for the cloud. The key objective of this study is to compute different task sizes in the Work flowSim cloud simulation environment using scheduling algorithms Max – Min (MxMn), Minimum Completion Time (M.C.T.), and Min – Min (MnMn) and later compare their behavior concerning various performance metrics. The exper imental results show that all these algorithms have dissimilar behavior, and no method supplies the best results under all performance metrics. Therefore, an in telligence mechanism is required to be provided to these resource scheduling al gorithms so they can perform better for all the performance metrics. Lastly, it is suggested that Reinforcement Learning (RL) acts as an intelligence mechanism, enhances the resource scheduling procedure, and makes the entire process dynamic, enhancing cloud performance.

Prathamesh Vijay Lahande, Parag Ravikant Kaveri
A Recommendation Model System Using Health Aware- Krill Herd Optimization that Develops Food Habits and Retains Physical Fitness

Major contributors to a healthy lifespan are regular physical activity and a nutritious diet. Although ageing is linked to decreased muscular function and endurance, these factors can be positively altered by regular physical activity, especially exercise training at all ages and in both sexes. Additionally, older people's altered body composition and metabolism, which can happen to even highly trained athletes, can be somewhat offset by active people's increased exercise metabolic efficiency. As a result, sportsmen are frequently cited as examples of good ageing, and their outstanding physical ability serves as a proof to what is still achievable as people age. This is due to their physical activities and the intake of food with high nutrient content. This paper provides a nutrient recommendation to the individual to improve their food habits and maintain the physical condition using the proposed Recommendation system using the Health Aware- Krill Herd Optimization (Recsys-HA-KHO). The Resys utilize HA-KHO to provide an efficient recommendation to consume nutritious food based on their Physical Activity (PA). The results obtained from the proposed HA-KHO is compared with other optimization algorithms based on the parameters likeprecision, recall, F1 score, Area Under Curve (AUC) and Discounted Cumulative Gain (DCG). The results show that the proposed HA-KHO obtained highest precision of 93.54% and the recall value of 97.32%.

N. Valliammal, A. Rathna
Video Summarization on E-Sport

Video summarizing is a useful technique for extracting the most important information from long videos. The “Video Summarization on E-sport” technique, which hasn't been extensively used in the esports industry yet attempts to automatically create highlights from e-sport game footage. This program, which users may input an e-sport game into, can produce summarized movies that capture the most thrilling parts of the game without the need for specialized equipment or experienced video editors. The suggested approach leverages AI, machine learning, and certain Python modules to produce highlights automatically, as opposed to the conventional way of making summarized movies, which is labor and time intensive. Users may save important time and effort by using the technology to automatically summarize key game moments by analyzing audio and text cues. The future of e-sports content creation may be greatly impacted by this technology, which has the potential to revolutionize the way e-sport game footage are summarized.

Vani Vasudevan, M. R. Darshan, J. V. S. S. Pavan Kumar, Saiel K. Gaonkar, Tallaka Ekeswar Reddy
SQL Injection Attack Detection and Prevention Based on Manipulating the SQL Query Input Attributes

SQL injection refers to one of the types of database attacks for web applications. The database security is compromised when wild card characters, malicious code, or malicious SQL query string are injected into the database. These changes in syntax and semantic allow the attacker to gain access to sensitive information and manipulate the database. Various techniques have been developed to detect and prevent this type of attacks. In this article, we proposed an method for preventing and detecting SQL injection. This method manipulates the SQL query input parameters and determining the distance between query strings. This method satisfies static query and dynamic also.

R. Mahesh, Samuel Chellathurai, Meyyappan Thirunavukkarasu, Pandiselvam Raman
Comparative Analysis of State-of-the-Art Face Recognition Models: FaceNet, ArcFace, and OpenFace Using Image Classification Metrics

In recent years, facial recognition has emerged as a key technological advancement with numerous useful applications in numerous industries. FaceNet, ArcFace, and OpenFace are three widely used techniques for facial identification. In this study, we examined the accuracy, speed, and capacity to manage variations in face expression, illumination, and occlusion of these three approaches over a period of five years, from 2018 to 2023. According to our findings, FaceNet is more accurate than ArcFace and OpenFace, even under difficult circumstances like shifting lighting and facial occlusion. Also, during the previous five years, FaceNet has shown a significant improvement in performance. Even while ArcFace and OpenFace have made significant strides, they still lag behind FaceNet in terms of accuracy. Therefore, based on our findings, we conclude that FaceNet is the most effective method for facial recognition and is well-suited for use in high-stakes applications where accuracy is crucial.

Joseph K. Iype, Shoney Sebastian
Hash Edward Curve Signcryption for Secure Big Data Transmission

Medical health service data is kept in huge dataset and shared through different devices. However, privacy and security of data sharing area significant concern, since data needs to be accessed from various locations in the distributed system. Therefore, a novel method called Hash Edward Curve Signcryption (HECS) is introduced for healthcare data communication in a secure manner. In the HECS method, related medical health service data is collected from the huge healthcare dataset. The proposed HECS method includes two major processes. First, Theil-Sen Robust Linear Regression is carried out to classifying the data and categorizes the data into number of classes. After the classification, the Hash Edward signcryption cryptographic technique is employed to perform the secured data transmission. In the designed cryptosystem, Pseudo ephemeral agreement key pair generation is performed for each session of data transmission. Then the HECS method performs the signcryption process for converting the data point into the encrypted data and generates the digital signature. Then in unsigncryption, the signature verification is carried out to check the user authenticity. This, in turn, secured data transmission is carried out with higher confidentiality. The assessment of the HESC method is carried with accuracy on classification and integrity of data. The quantitatively analysed results confirmed that the HECS method increases security than the orthodox methodology by means of improved data confidentiality rate and integrity rate.

S. Sangeetha, P. Suresh Babu
An Energy Efficient, Spontaneous, Multi-path Data Routing Algorithm with Private Key Creation for Heterogeneous Network

In the recent years many investigation studies are carried out in heterogeneous sensor networks owing to its countless applications in all fields like health care, defense, environment surveillance, industrial based applications etc. In general, heterogeneous network have many fruitful advantages as precise sensing of parameters, manipulation of sensed data and passing sensed dataset to a receiver point in an effective manner. This type of network may be incorporated in a large-scale environment in an unstructured network with energy limitations. As known that sensor has a defined amount of power, memory and coverage area. For realistic implementation of nodal point in the investigation premises, it should be defined in an effective manner both in physical and coverage metrics. Hence it is unavoidable to propose a methodology for the effective usage of power conservation to achieve an optimistic performance. Several research studies were carried out in former years based on Fuzzy technique, Neural Networks, genetic-algorithm and many more improved approaches. This paper proposes a Population based (Ant Colony Optimization-ACO) investigative approach for a multi-path data forwarding to heterogeneous network as spontaneous energy proficient multi-path data routing (SEPMDR). In the initial stage, the proposed approach discovered its efficient neighboring sensor nodal before forwarding data packets with improvised data security factor. The privacy in the data transmission is implemented by using security key generation for each data transmission carried out in the heterogeneous network. The stability of the proposed technique in terms various network parameters with previously defined research strategies is evaluated and simulation results are conferred with Network Simulator Tool.

K. E. Hemapriya, S. Saraswathi
A Hybrid Model for Epileptic Seizure Prediction Using EEG Data

More than 65 million people’s quality of life is affected by a neurological brain condition epilepsy. When a seizure can be anticipated, therapeutic action can be employed to stop it from happening. The data analysis of epileptic seizures makes use of EEG impulses. In the interim, amplitude integrated Electroencephalography (aEEG) has shown promise in the identification of epileptic episodes. This article describes a method for automatically identifying epileptic episodes in EEG readings. Three steps make up the suggested methodology: preprocessing, feature selection, and classification. The current study proposes a deep learning-based seizure prediction system that includes preprocessing scalp EEG signals, extracting key characteristics implementing convolutional neural networks (CNN), and classifying them with the as distance of vector machines. The put forward framework demonstrates its powerful capability in the automatic the identification of seizures, as evidenced by its competitive hypothetical results on EEG datasets analyzed to state-of-the-art approach with a success rate of 98.57%.

P. S. Tejashwini, L. Sahana, J. Thriveni, K. R. Venugopal
Adapting to Noise in Forensic Speaker Verification Using GMM-UBM I-Vector Method in High-Noise Backgrounds

The performance of the GMM-UBM-I vector in a forensic speaker verification system has been examined in the context of noisy speech samples. This analysis utilised both Mel-frequency cepstral coefficients (MFCC) and MFCCs generated from auto-correlated speech signals. The noisy signal’s auto correlation coefficients are concentrated around the lower lag, whereas the autocorrelation coefficients near the higher lag are very small. Thus, in addition to retain the periodic nature, autocorrelation-based MFCC is also robust for analyzing speech signals in intense background noise. The performance of MFCC and auto-correlated MFCC depends heavily on the quality of the sample. It works best with data that is free of noise, but it suffers when used on real-world examples, ie, with noisy data. The experiment on speaker verification for forensic purposes involved the addition of White Gaussian Noise, Red Noise, and Pink Noise, with a Signal-to-Noise Ratio (SNR) range spanning from −20 dB to + 20 dB. The performance of both methods was affected drastically in call cases but autocorrelation-based MFCC gave better results than MFCC. Thus, autocorrelation-based MFCC is a valuable method for robust feature extraction when compared with MFCC for speaker verification purposes in intense background noise. The verification accuracy in our method is improved even in very high noise levels (−20 dB) than the reported research work.

K. V. Aljinu Khadar, R. K. Sunil Kumar, N. S. Sreekanth
Classification of Code-Mixed Tamil Text Using Deep Learning Algorithms

Natural Language Processing (NLP) is a vast subject with applications in many fields in today’s modern world. The goal of NLP is to achieve human like language processing for a variety of activities or applications. The internet is full of textual data in many different languages. Although a large number of internet comments found in public spaces are often positive, a significant portion are toxic in nature. We first need to separate the good from the bad before classifying the different levels of toxicity. This will lessen any unintentional prejudice towards certain individuals or entity and lessen negativity on social media. Our primary goal is to identify, categorize, and analyze the toxicity that now plagues social media platforms. This study focuses on classifying Code Mixed Tamil text using deep learning algorithms. Tamil as a language has many obstacles to be overcome in this NLP task. Since Tamil’s grammar structure, specific features are unique and complex, it is actually hard to make a model that can consistently perform for any data from the language of Tamil. The agglutinative nature of Tamil is a major problem when it comes to tasks like classification since the context gets twisted when the single word is split into corresponding morphemes. Since there are many studies conducted on Code Mixed text of other languages with deep learning algorithms, this paper aims to find the effectiveness of XLNet and Bi-LSTM on Code-Mixed Tamil.

R. Theninpan, P. Valarmathi
The Road to Reducing Vehicle CO2 Emissions: A Comprehensive Data Analysis

In recent years, the influence of carbon dioxide (CO2) releases on the environment have become a major concern. Vehicles are one of the major sources of CO2 emissions, and their contribution to climate change cannot be ignored. This research paper aims to investigate the CO2 emissions of vehicles and compare them with different types of engines, fuel types, and vehicle models.The study was carried out by gathering information about the CO2 emissions of vehicles from the official open data website of the Canadian government. Data from a 7-year period are included in the dataset, which is a compiled version. There is a total of 220 cases and 9 variables. The data is analyzed using statistical methods and tests to identify the significant differences in CO2 emissions among different Car Models. The results indicate that vehicles with diesel engines emit higher levels of CO2 compared to those with gasoline engines. Electric vehicles, on the other hand, have zero CO2 emissions, making them the most environmentally friendly option. Furthermore, the study found that the CO2 emissions of vehicles vary depending on the type of fuel used. The study also reveals that the CO2 emissions of vehicles depend on the model and age of the vehicle. Newer models tend to emit lower levels of CO2 compared to older models. In conclusion, this study provides valuable insights into the CO2 emissions of Cars and highlights the need to adopt cleaner and more sustainable transportation options.

S. Madhurima, Joseph Mathew Mannooparambil, Kukatlapalli Pradeep Kumar
A Deep Learning Based Bio Fertilizer Recommendation Model Based on Chlorophyll Content for Paddy Leaves

Rice is the main agricultural product in India, with 90% of the population consuming it as a staple food. Nutrient deficiencies in rice leaves during growth period causes imbalances leading to reduce in crop yield. Growth of the paddy plants is highly related to the chlorophyll content present in it. Chlorophyll content in a leaf is a key indicator of greenness of a leaf and identifies the nutrient deficiencies in plants. Chlorophyll in plants contains nitrogen which is responsible for the photosynthesis process. In this research, chlorophyll and nitrogen contents are measured for the paddy leaves. Here, Support Vector Machine (SVM) Regression and Convolutional Neural Network (CNN) models are used to measure the chlorophyll and nitrogen content in plants. In this research, the color was the main parameter used to quantify chlorophyll and nitrogen contents in plants and RGB color model is used. Bio fertilizers is an influencing factor in yield progression and physiological processes as Bio fertilizers supply necessary nutrients to plants and enhance chlorophyll content in leaves. The chlorophyll and nitrogen concentrations were measured and then based on the measured nitrogen concentration, the appropriate bio fertilizer is recommended by the SVM classification model to enhance the nitrogen content in the plants. Here the CNN (99.98%) algorithm works better than the SVM algorithm in prediction of Chlorophyll and nitrogen contents. Hence, the Convolutional Neural Network model is built to predict the chlorophyll and nitrogen contents for the paddy leaves based on color and recommends the appropriate bio fertilizer to improve plant growth.

M. Nirmala Devi, M. Siva Kumar, B. Subbulakshmi, T. Uma Maheswari, Karpagam, M. Vasanth Kumar
A Comparison of Multinomial Naïve Bayes and Bidirectional LSTM for Emotion Detection

Emotion detection is an area of sentiment analysis that focuses on the extraction and evaluation of feelings. Many deep learning and machine learning researchers have found emotional content in text. People’s lives all across the world were significantly impacted by the COVID-19 pandemic. The social networking site twitter were very helpful in documenting people’s emotion and views. In literature there are two methods that has been used extensively for the emotion identification; one is Multinomial Naive Bayes model, and another is Bidirectional Long Short-Term Memory (LSTM). In this research, we have identified the efficient approach among the two mentioned approaches. We have compared the multinomial naive bayes model and bidirectional LSTM for identifying emotion. To identify the emotion in text, the tweet from twitter is utilized; that contains emotions in the form of text. The results show that bidirectional Long-Short Term Memory approach outperforms as compared to the multinomial naïve bayes approach (MNB). The bidirectional LSTM approach enhances the emotion detection over the MNB approach. The time taken for the training of the tweets differs for the two approaches. Due to its longer training period, the multinomial naive bayes technique may not be suitable for use with huge datasets.

S. K. Lakshitha, V. Naga Pranava Shashank, Richa, Shivani Gupta
Hybrid Region of Interest Based Near-Lossless Codec for Brain Tumour Images Using Convolutional Autoencoder

One of the most significant industries producing digital images worldwide is radiology. The advancements in radiological equipment have ensured the production of high-definition digital medical images. Irrespective of the growth in storage and network facilities, these high-definition images still face the problem of high storage space requirements and high transmission costs. To handle aforementioned problems and pave way for faster, hassle-free transmission and effective storage of such medical images for telemedicine services, we need enhanced image compression techniques. Medical images tend to have data with both high and low clinical importance for downstream analysis and treatment. Compression algorithms must be developed in order to handle both high and low clinically important data at the same time to improve the compression standard of medical images. We propose a Convolutional autoencoder technique for Region of Interest based hybrid near-lossless medical image compression aided by “You Only Look Once” (YOLO) deep learning algorithm. This work aims to achieve ROI-based near-lossless compression with notable compression ratio and medical image quality. To achieve this ROI-based near-lossless compression, we employed a combination of YOLO object detection algorithm, Convolutional autoencoder, and Haar wavelet transform with SPIHT encoding on grayscale Magnetic Resonance brain tumour images. The proposed approach was evaluated against several existing standard compression methods. Results inferred that our proposed method assured the near-lossless image compression scenario by maintaining the quality of medical images after decompression and comparatively reduced the storage and transmission cost by ensuring an effective compression ratio.

Muthalaguraja Venugopal, Kalavathi Palanisamy
An Empirical and Statistical Analysis of Classification Algorithms Used in Heart Attack Forecasting

The risk of dying from a heart attack is high everywhere in the world. This is based on the fact that every forty seconds, someone dies from a myocardial infarction. In this paper, heart attack is predicted with the help of dataset sourced from UCI Machine Learning Repository. The dataset analyses 13 attributes of 303 patients. The categorization method of Data Mining helps predict if a person will have a heart attack based on how they live their lives. An empirical and statistical analysis of different classification methods like the Support Vector Machine (SVM) Algorithm, Random Forest (RF) Algorithm, K-Nearest Neighbour (KNN) Algorithm, Logistic Regression (LR) Algorithm, and Decision Tree (DT) Algorithm is used as classifiers for effective prediction of the disease. The research study showed classification accuracy of 90% using KNN Algorithm.

Gifty Roy, Reshma Rachel Cherish, Boppuru Rudra Prathap
Healthcare Data Analysis and Secure Storage in Edge Cloud Module with Blockchain Federated Sparse Convolutional Network++

In current scenario, most difficult requirements are managing massive amount of multimedia data generated by Internet of Things (IoT) devices, which can only be managed with the cloud. The intelligent Edge Cloud computing technology operates in a distributed environment and emerges as a solution. This research aims to use Edge Cloud computing to reduce latency in e-healthcare. This study proposes a novel method for using machine learning to analyze healthcare data and storing it in an edge cloud module. The input is monitored healthcare data that is collected, processed, and analyzed with the help of a blockchain-federated sparse convolutional network++, which also improves network security. The malicious attacks are then identified and malicious data is stored using a centralized edge cloud computing module and authentication process. The experimental analysis is conducted on a variety of healthcare datasets and is based on a security analysis of the data in terms of data transmission rate, computation cost, communication overhead, random accuracy, mean average precision (map), and specificity. The proposed method achieved a data transmission rate of 65%, a computation cost of 51%, a communication overhead of 75%, random accuracy of 85%, mean average precision (map) of 63%, and specificity of 79%.

R. Krishnamoorthy, K. P. Kaliyamurthie
A Multi-layered Approach to Brain Tumor Classification Using VDC-12

This research paper presents a deep convolutional neural network (CNN) model approach along with feature extraction for multiclass brain tumor detection using medical imaging. Compared to traditional methods that rely on manual interpretation of scans, the VDC-12 (Very Deep Convolution) model proposed in this study has the potential to enhance the accuracy of detecting brain tumors. We used a dataset of MRI brain images containing four categories of tumors, namely meningioma, glioma, pituitary, and normal brain tissue. To evaluate the proposed CNN model’s performance, various metrics were used, including accuracy, precision, recall, and F1 score. According to the experimental results, the VDC-12 model surpasses several cutting-edge techniques, achieving a classification accuracy of 97.60%. This model exhibits promise in the early detection of brain tumors, which can facilitate timely diagnosis and treatment.

Anant Mehta, Prajit Sengupta, Prashant Singh Rana
An Efficient Approach of Heart Disease Diagnosis Using Modified Principal Component Analysis (M-PCA)

Heart diseases have come to be a first-rate purpose of demise around the arena. As a end result, heart ailment prediction has obtained plenty of interest in the medical global. As a end result, several studies have developed machine-getting to know algorithms for the early prediction of coronary heart sicknesses to help physicians inside the design of clinical processes. The performance of those structures is determined largely by way of the function set decided on. When the schooling dataset consists of missing values for the distinct capabilities, this will become greater hard. The opportunity of Principal Component Analysis (PCA) to solve the trouble of lacking attribute values is widely known. This studies presents a technique for diagnosing heart sickness via taking scientific checking out results as enter, extracting a low dimensional characteristic subset, and diagnosing coronary heart sickness. Modified Principal Component Analysis (M-PCA) is used within the proposed method to extract better depth features in new projections. PCA aids within the extraction of projection vectors that make a contribution substantially to the maximum covariance and uses them to lessen function size. The proposed method is analysed across three datasets, and the effects, accuracy, sensitivity, and specificity are calculated. To illustrate the implications of the proposed M-PCA technique, the received results the use of it are in comparison to previous research. The proposed M-PCA technique produced an extremely correct dataset.

G. Lakshmi, P. Sujatha
Smart Driving Assistance Using Deep Learning

This research paper presents a smart driving assistance system that utilizes deep learning for lane detection and departure alert, traffic sign detection, and voice alerts. The system uses a combination of computer vision techniques and neural networks to detect lanes and traffic signs in real-time. The lane departure alert feature alerts the driver when the vehicle begins to drift out of its lane, and the traffic sign detection feature identifies and alerts the driver of any traffic signs that are relevant to the current road. The voice alert feature provides an additional layer of safety by audibly alerting the driver of any detected traffic signs. The proposed system has been evaluated on a dataset of real-world driving scenarios and has shown promising results in terms of accuracy and efficiency.

S. N. Baba Shankar, B. Karthik Reddy, B. Koushik Reddy, Venuthurla Venkata Pradeep Reddy, H. B. Mahesh
Design of Advanced High-Performance Bus Master to Access a SRAM Block

The Advanced Microcontroller Bus Architecture (AMBA) is an open System-on-Chip bus protocol for high performance buses to communicate with low-power devices. In the AMBA Advanced High-performance Bus (AHB), a system bus is used to connect a processor, direct memory access (DMA), and high-performance memory controllers. The AMBA Advanced Peripheral Bus (APB) is used to connect UART (Universal Asynchronous Receiver Transmitter). It also contains a bridge, which connects the AHB and APB buses. Bridges are standard bus-to-bus interfaces that allow Interface Protocols connected to different buses to communicate with each other in a standardized way. In this work, we have done an interface between the AHB and SRAM which performs read and write operation. The AHB SRAM Controller provides a standard AHB interface to translate AHB bus reads and writes into reads and writes with the signaling and timing of a standard 32-bit synchronous SRAM. The design has been simulated in EDA Playground with Riviera-pro tool and EPWave to display the output.

M. S. Mallikarjunaswamy, Jagadeesh Dambal, Amish Kuthethure, G. Ashwini, K. Sumanth
EFMD-DCNN: Efficient Face Mask Detection Model in Street Camera Using Double CNN

The COVID-19 pandemic has necessitated the widespread use of masks, and in India, mask-wearing in public gatherings has become mandatory, with violators being fined. In densely populated nations like India, strict regulations must be established and enforced to mitigate the pandemic’s impact. Authorities and cameras conduct real-time monitoring of individuals leaving their homes, but 24/7 surveillance by humans is not feasible. A suggested approach to resolve this problem is to connect human intelligence and Artificial Intelligence (AI) by employing two Machine Learning (ML) models to recognize people who aren’t wearing masks in live-stream feeds from surveillance, street, and new IP mask recognition cameras. The effectiveness of this method has been demonstrated through its high accuracy compared to other algorithms. The first ML model uses the YOLO (You Only Look Once) model to recognize human faces in real-time video streams. The second ML model is a pre-trained classifier using 180,000 photos to categorize photos of humans into two groups: masked and unmasked. Double is a model that combines face recognition and mask classification into a single model. CNN provides a potential solution that may be utilized with image or video-capturing equipment such as CCTV cameras to monitor security breaches, encourage mask usage, and promote a secure workplace. This study’s proposed mask detection technology utilized pre-trained datasets, face detection, and various classifiers to classify faces as having a proper mask, an improper mask, or no mask. The Double CNN-based model incorporated dual convolutional neural networks and a technology-based warning system to provide real-time facial identification detection. The ML model achieved high performance and accuracy of 98.15%, with the highest precision and recall, and can be used worldwide due to its cost-effectiveness. Overall, the proposed mask detection approach can potentially be a valuable instrument for preventing the spread of infectious diseases.

R. Thamarai Selvi, N. Arulkumar, Gobi Ramasamy
Reflecting on Technology: A Review of the Smart Mirror Advancements and Applications

Smart mirrors gained popularity in recent years due to their unique ability to enhance our daily routines. They have been designed to improve the user experience of using a mirror and have the potential to provide a range of services, including displaying weather forecasts, news updates, time, social media notifications, fitness information etc. Smart mirrors have many potential applications which can enhance daily routines and improve quality of life. In this paper, the recent advancements and applications of smart mirrors are reviewed.

Divyesh Divakar, H. M. Bharat, G. M. Dhanush
MobNetCov19: Detection of COVID-19 Using MobileNetV2 Architecture for Multi-mode Images

COVID-19 created a history in the world of medicine which leads to more usage of technologies such as deep-learning models to aid in the early detection of COVID-19 using medical imaging from three commonly used modalities: X-Ray, Ultrasound and Computerized Tomography (CT) scan. This research aims to provide medical professionals with an additional tool to assist in devising an appropriate treatment plan and making disease containment decisions. We have identified the suitable optimized VGG19 and MobNetCov19 architecture through a Convolutional Neural Network (CNN) model for a comparative study of the different imaging modes to develop highly curated COVID-19 detection models despite the scarcity of COVID-19 datasets. Our results demonstrate that CT dataset has the highest detection accuracy compared to X-Ray and Ultrasound datas. Although the limited data made training complex models challenging, the selected MobNetCov19 model, extensively tuned with appropriate parameters, performed considerably well up to 100%, 98%, and 98% of accuracy for CT, X-Ray, and Ultra sound respectively.

H. S. Suresh Kumar, S. Bhoomika, C. N. Pushpa, J. Thriveni, K. R. Venugopal
Analysis of Machine Learning Approaches to Detect Pedestrian Under Different Scale Using Frame Level Difference Feature

Over the last 20 years, automotive technology has advanced to the point where automated systems can currently handle various aspects of vehicle control. Due to traffic congestion, pedestrians are particularly vulnerable and they collide with the vehicle’s front end. In recent years, the legal standards and consumer protection assessments for pedestrian protection have gotten much stronger. Sensor technology that must reliably detect an impact between a vehicle and a person has substantial hurdles as a result. Computer vision-based technologies play an important role in the enhancement strategies of automation industries like the Advanced Driver Assistant System (ADAS) by identifying and tracking people on the road. During the process of pedestrian identification, human characteristics are a key factor in determining accuracy. The extraction of features to identify the pedestrians from the video images is a difficult task. In this paper, a per-frame evaluation methodology is taken in hand to make in depth and insightful comparisons among state of art detection techniques with FLD features. This investigates the detection rates at different scales. The experiments that are conducted with the help of Caltech pedestrian dataset. The recognition accuracy is compared and evaluated by avoiding a higher number of erroneous hits and a higher percentage of miss rates.

A. Sumi, T. Santha
Eye Support: AI Enabled Support System for Visually Impaired Senior Citizens

Individuals who have impaired eyesight have a difficult time seeing the tablet’s name, which is written in small characters on the back of the tablet. India features in the “top ten” list for the visually impaired. Numerous technologies are growing in India, allowing us to aid the populace in identifying the tablet’s name, purpose, and other information without assistance from others. Increasingly, technology is being employed for a variety of reasons, and individuals are employing smart gadgets and devices. In the suggested method, users may launch the app on their smartphones, upload or snap a photograph of the pharmaceutical strip, and receive the findings in audio format. The proposed system employs Artificial Intelligence (AI), and Natural Language Processing (NLP) in Python for processing and output in voice format. The smartphone application primarily benefits the elderly, notably those with visual impairments.

Vani Vasudevan, Thota Thanmai, Riya Yadav, Pola Udaya Sowjanya Reddy, Subrina Pradhan
The Evolution of Block Chain Technology

Block chain is the latest trending technology that is used in many fields worldwide. Block chain technology has been implemented in many leading areas which is brought up to date on in this paper. Its applications consist of concepts that can be traced and easily tracked in the form of blocks by using a Blockchain network. These blocks acts like a storage for data to perform transactions which are further chained together by validating and adding the block chaining the previous hash address. The main objective of this paper is to gain deep knowledge about the implementation of Blockchain technology, the areas of application, the working of Blockchain technology by incorporating different algorithms like Proof of Stake, Delegated Proof of Stake, Proof of Authority, Proof of Burn etc. and also highlight about its concepts, types, and the challenges of the Block chain technology. It also gathered many core concepts implemented in the research.

Indumathi Karthikeyan, M. G. Shruthi
Computational Sciences and Sustainable Technologies
herausgegeben von
Sagaya Aurelia
Chandra J.
Ashok Immanuel
Joseph Mani
Vijaya Padmanabha
Electronic ISBN
Print ISBN

Premium Partner