This exhaustive investigation examines the Advanced Driver Assistance System (ADAS) in driverless cars. ADAS is vital to improve the safety and effectiveness of an autonomous car. The research briefs the integration of ADAS components such as sensors, cameras, and radar systems into the vehicle's control system. The work also examines the obstacles associated with the implementation of ADAS technology in driverless cars, including data management, cyber-security, and regulatory requirements. In addition, the study evaluates the advantages and disadvantages of ADAS technology and its influence on the automobile industry's future. The results indicate that although ADAS technology provides considerable benefits, such as enhanced safety and driving enjoyment, there are also potential dangers connected with the usage of autonomous cars that must be addressed.
As a deep feed-forward neural network, the convolutional neural network (CNN) model has achieved major breakthroughs in scheme of image recognition. Compared with the traditional classification methods (KNN, SVM, PSO), the fully automatic classification algorithms reply on convolutional network is separated from human identification, which is tedious, and at the same time backpropagation algorithm is used to automatically optimize the model parameters. Inspired by this, the paper proposes the novel intelligent image recognition scheme with the integrations of fully convolutional neural network. (FCNN). To solve the challenge that overall training accuracy of the general model is reduced due to the overfitting phenomenon caused by the commonality of patterns of the image features while training the models, this study introduced the AlexNet. The loss function is improved and optimized, and for preprocessing of the images, the compressive analysis model is integrated. Through the single and comparison tests, the performance of the designed algorithm is tested, and overall speaking, the recognition accuracy is better than the traditional methodologies.
The most crucial components of life is health. People could still not access adequate healthcare, nevertheless. It is brought on by restrictions on the technology used in hospitals and on how people may travel to hospitals. One of today's most popular topics, the Internet of Things (IoT), already provides many solutions, including in the healthcare industry. The field of wearable medical technology has emerged as one of the promising healthcare technologies. By actively collecting physiological indicators and monitoring metabolic status, these intelligent sensors help individuals lead healthier lifestyle and provide continuous healthcare information for disease diagnosis and treatment. Wearable technology is a new technique to track patients' post-intervention progress. It may lead to better patient outcomes and less demand for healthcare resources by enabling a faster and safer discharge from the hospital. Current findings have looked into the usage of wearable technology for monitoring. The ability to incorporate gadgets that can connect to the internet, provide data on patient health, and give real-time data to doctors who can help is becoming more and more possible due to the Internet of Things (IoT). It is obvious that health indicators like temperature, heartbeat, high blood pressure, and humidity represents a significant global economic and social issue. The patient's health is being monitored for health issues, and details can be viewed for obtaining the live patient data on a mobile web server at the correct time.
Sentiment analysis, a subset of Natural Language Processing (NLP), has grown in relevance as the modern lifestyles have become more reliant on social media. Emotional expressions include actions, feelings, ideas, and nonverbal cues. Modern approaches for Natural Language Processing (NLP) have their origins in machine learning, notably statistical machine learning. This study introduces BERT, an innovative technique for reading sentiments into written language. This model is a synthesis of motor Bidirectional Encoder Representations From Transformers (BERT) representations. In order to develop its learning algorithm of word semantic representation, this model makes use of the BERT. In line with the language context, a comparison of models found that BERT outperforms the government foundation efficiency given by all the other models in the literature. Opinion mining is a branch of natural language processing that analyses public discourse to gauge public opinion on a product or topic.
Due to the possibility of cyberattacks, cybersecurity has emerged as one of the most crucial issues relating to IoT technology. IoT cybersecurity aims to lessen cybersecurity hazards for businesses and customers by safeguarding IoT resources and privacy protection. New cybersecurity technologies and strategies may make it possible to handle IoT security better. Internet-connected items such as smartphones, smart schooling, smart transportation and smart cities were all made possible by modern technology. The most significant application area for ML-based strategies to handle frequent attack challenges and create thought-provoking conversations is thought to be smart transportation. An efficient BoT-IoT dataset that already exists is employed for this purpose, together with a variety of attack categories and subcategories, for training and evaluating the system’s dependability. Using the BoT-IoT dataset, the paper’s primary objective is to deploy various machine learning techniques, including Random Forest (RF), Naive Bayes (NB), and Decision Tree (DT), to analyse the effectiveness of attacks. Using the most well-known BoT-IoT dataset, the best accuracy obtained by the machine learning methods, RF and DT is 91% and 91%, respectively.
A comparator is a key component in variety of applications such as data converters, mixed signal systems, signal processing, digital I/O circuits, biomedical devices, memory sensing circuits, and so on. These applications frequently necessitate low-power, high-speed and low-offset comparators. In particular for signal processing applications, enhancing the performance of data converters depends on the accuracy of comparators. The most effective way to enhance the sampling rate of data converters is to improve comparator speed. A pre-amplifier with a charge pump is used in the proposed dynamic comparator, followed by a decision circuit and an output stage. The simulated results demonstrate that comparator has achieved a low power consumption of 13.13μW with a delay of 85.54ps making it more suitable for data converters. The comparator operates with clock frequency of 1GHz and a supply voltage of 1V, and it is simulated in 45nm CMOS technology using the Cadence virtuoso tool.
In recent years, the increasing demand for secure image transmission and storage has necessitated the development of robust and efficient image encryption algorithms. This study presents a novel color image encryption algorithm based on an 8-dimensional (8D) hyperchaotic system and DNA encoding techniques. The proposed algorithm exploits the inherent complexity and unpredictability of the 8D hyperchaotic system to enhance security, while leveraging the unique properties of DNA sequences for encoding and data manipulation. The proposed encryption process consists of three main stages: (1) generation of chaotic sequences from the 8D hyperchaotic system, (2) DNA encoding of the color image using a set of predefined DNA encoding rules, and (3) an encryption process that combines chaotic sequences with DNA-encoded image data through a series of substitution and permutation operations. The decryption process reverses these stages to recover the original image. Comprehensive security analysis, including key space analysis, sensitivity analysis, and correlation analysis, demonstrates that the proposed algorithm provides a high level of security against various attacks. Additionally, performance evaluation based on image quality metrics, such as Mean Square Error (MSE) and Peak Signal-to-Noise Ratio (PSNR), confirms the efficiency and effectiveness of the proposed algorithm for color image encryption applications.
This research survey presents a comprehensive analysis of IoT architectures that incorporate edge computing, examining their classifications, applications, and the challenges and limitations associated with their implementation. The survey also explores future directions and trends in IoT architectures that leverage edge computing, including 5G integration, artificial intelligence (AI), and machine learning (ML) advancements, edge virtualization, edge security, and improved interoperability. The research study highlights the potential of IoT architectures that utilize edge computing to address the challenges posed by traditional cloud-centric architectures while also acknowledging the key challenges and limitations that need to be overcome. By examining the current state of the field and emerging trends and technologies, this survey aims to provide a valuable reference for researchers, practitioners, and industry stakeholders working on the development and implementation of IoT solutions incorporating edge computing.
Direction of arrival (DOA) estimation in 5G massive MIMO mm Wave systems, is crucial for effective beamforming. A high-resolution DOA estimate is required to avoid interferences and steer the beam to the corresponding user. The accuracy of DOA estimation deteriorates significantly due to the antenna array imperfections, channel fluctuations, and noise. This research investigates the performance of DOA estimation with a uniform linear array of 128 elements at a 5G Frequency Range FR2 of 26GHz and a single antenna user. The antenna array imperfections considered are antenna position perturbations, inconsistent gain, and phases. To cope with these array imperfections, the pre-processing technique is done by evaluating the compensation matrix with the help of the Least Square (LS) approach and Orthogonal Matching Pursuit (OMP). The compensation matrices are estimated with the steering vectors of ideal and practical array manifolds. Then the received signals are compensated for imperfections with the compensation matrix. The compensated signals are then fed into the DOA estimator, and the performance analysis of Multiple Signal Classifier (MUSIC), Estimation of Signal Parameters via Rotational Invariance Technique (ESPRIT), Root-MUSIC (RMUSIC), and Beam Scan algorithms are compared. The quantitative results and analytical formulation of the estimation technique are discussed based on the results of the DOA estimation techniques. Simulation results show that resolution of the DOA estimation of Root-MUSIC and ESPRIT with OMP achieves better performance-complexity trade-off at low SNR region, whereas MUSIC and Beam scan achieves better performance-complexity trade-off at high SNR region than LS compensation technique.
In this modern era, the regular life of people have become more competent and are interlinked with technology. There are many voice assistants like Google and Siri. etc., getting used in the present day. Here, the proposed Artificial Intelligence Enabled Voice Assistance System (ARIVA) using Natural Language Processing facilitates domains such as drug prescribers, to-do lists, note-writing, calculators, and searching tools. The ARIVA model takes input as a voice signal and yields output in numerous ways, like voice and a visual display. This virtual assistant aims to provide users with instant, accomplished results. The proposed ARIVA system accepts the command via the microphone of the executing system and converts human voice into computer-understandable language, and then generates the required solutions for the said queries. The speech recognition and processing are carried out using Natural Language Processing (NLP) algorithm that aids computer systems in interacting through natural human speech in different forms. ARIVA can connect with the world wide web to procure the results of the individual's query. The ARIVA aids the user with everyday actions such as general verbal interaction, exploring Google for query processing, video searching, searching for videos, playing songs, predicting future weather reports, analyzing synonyms, searching for health and hygiene, and assisting the user by scheduling tasks and events. In terms of performance, ARIVA, when compared to IntelliAssistant, it is seen that ARIVA consumes lesser time and exhibits lesser memory requirement.
Roadside potholes raise maintenance costs for road officials while causing serious harm to vehicle and endangering the safety of drivers and passengers. In this study, we use Convolutional Neural Network (CNN) technology to suggest a complete system for automatic pothole recognition and vehicle speed management. A pothole detection module and a vehicle speed control module are the two major parts of our system. A CNN-based algorithm is used by the pothole recognition module to evaluate the images. The model is trained to recognize potholes and identify them apart from other roadside characteristics. The vehicle speed control mechanism receives an information when a pothole is found. On a dataset of actual road images, the proposed model is tested and the potholes are identified with an accuracy of 99.56%. The proposed system provides a useful and effective solution for pothole recognition and vehicle speed control, which can help reduce accidents, save money on maintenance, and enhance the driving experience in general.
Lung diseases are the leading causes of early death and disability worldwide. Each year, they kill 4 million people and make them unable to work. One of the pulmonary diseases is emphysema. Emphysema, caused by the breakdown of alveolar walls and a lack of elasticity, is one of the most prevalent diseases responsible for this condition. Many algorithms have been developed in the past to classify emphysema. The objective of this article is to create a transfer learning-based model to determine whether a patient has emphysema based on chest X-rays. DenseNet201, a CNN classifier based on transfer learning, is proposed in this research. The suggested approach was tested using the NIH chest X-ray dataset (7540 samples). Finally, the experiments revealed that the DenseNet201 has a classification accuracy of 98.87%, an AUC of 99.7%, an F1-score of 98.88%, a precision of 97.80%, an AUC of 99.7%, a sensitivity of 100%, a precision of 97.80%, a validation accuracy of 98%, and a specificity of 97.75%, which is higher than other models. Other hyperparameters, such as learning rates, epochs, batch size, number of filters, and activation function, have been tuned repeatedly to improve results. The main goal is to have a low number of false positives and a 0% false negative rate.
Most businesses struggle with managing their finances and keeping track of their expenses. The current prototyping system presents design and implementation of an automatic billing and tracking system that can help businesses streamline their financial operations. The system is based on a web application that allows business owners to easily create and manage invoices, track expenses, and generate reports. This work offers a one stop solution for long wait in queues which is a cumbersome task. Weight sensors, RFID scanners are used to build a smart cart providing solution for various problems arising during shopping such as manual intervention while billing, miscounting of purchased goods, miscalculation of weights, thefts and other such discrepancy. With Internet of Things (IoT) principles in place, the offered prototype proves to be superior when compared to the existing smart carts which do not address the thefts thereby making the entire financial management flexible. The interface is made friendly for non-technical users as well.
The location of a sensor node is crucial for several uses of WSN, including environmental sensing, search and rescue, geographical routing and tracking, and so on. The accuracy with which individual sensor terminals in a wireless sensor system can be located has a substantial bearing on the network's overall effectiveness. Using information about the locations of anchor nodes gathered from a variety of measures to pinpoint the unknown target nodes' placements is known as localization. This is a problem classed as NP-hard, which means it cannot be solved using classical deterministic methods. To overcome this difficulty in wireless sensor networks, presented an enhanced version of a swarm intelligence technique called the whale optimization method. This implementation, using a quasi-reflected-based learning method, fixes the problems with the original whale optimization approach. In order to ensure that the proposed metaheuristic performs as well as existing state-of-the-art metaheuristics, it is evaluated using the same network architecture and experimental settings. Proposed method use a Gaussian-modified RSSI to achieve a more accurate reading of the range and a new whale optimization algorithm to optimize the positioning of the nodes to boost the positioning accuracy, both of which are designed to compensate for the shortcomings of the positioning algorithm of both Received signal strength indicator (RSSI) ranging model. Based on the results of 20 separate benchmark function tests, the upgraded whale algorithm outperforms the whale optimization method and other swarm intelligence systems. The suggested location algorithm provides more precise placement than the original RSSI method. It is mentioned that the cluster intelligence algorithm has significant benefits over the currently implemented RSSI in positioning WSN nodes, and the improved algorithm that is described in this work has even more benefits compared to various cluster intelligence methods in tends to work the locating needs of real-world applications. The developed approach, as shown by simulation results, achieves better localization accuracy than the baseline whale optimization technique and other leading metaheuristics.
Millions of patients suffer from epilepsy each year, a chronic nervous illness with a growing global prevalence. In a lot of situations, it might cause critical injuries or patient deaths. So, the automatic prediction of the epileptic seizure saves the patients from injury as well as from death. In this age of technological advancements, the internet of things (IoT) has started to offer a variety of solutions in the healthcare sector through the use of machine learning, deep learning, and cloud-based services. Internet of thing, Edge computing, and Cloud computing helps in the detection of such nervous disorders using deep learning and machine learning. Most of the models were developed preciously but there are many issues related to the prediction and security of the data of patients. This research study proposed an autonomous cloud-edge integrated seizure prediction model based on a convolutional neural network. The proposed model detects and classifying the epileptic seizure using an EEG dataset. The proposed model shows average training accuracy and average validation accuracy are 99.27% and 96.37%, respectively and the average losses for training and validation are 2.10 % and 14.23%, respectively. The result shows that overall accuracies and losses are better than the wellknown existing works. All implementation for seizure prediction is done in Python language.
For a long time, the Moroccan stock has attracted the attention of academics and speculative financiers alike. Sentiment analysis has developed into a potent instrument for foreseeing market movements, thanks to the growing importance of social media & the internet. With the use of ML & sentiment analysis methods, this research suggests a new way of looking at the Moroccan stock market. Using NLP, the study compile opinions from a wide range of internet sources, such as social media and news sites. Then Naive Bayes is used & other ML methods to make predictions about the correlation between public opinion & stock market outcomes. Proposed findings point to encouraging associations between mood and market patterns, suggesting that sentiment research has potential as a technique for foreseeing shifts in the stock market. Significant consequences for traders, financial experts, and policymakers interested in the Moroccan stock markets may be drawn from this research.
A quick and reliable security systems or APIs are required since the communication domain and interactive media data, such as movies and photographs, are expanding exponentially. Encryption is one method of protecting our data from unauthorized users when communicating and sharing files. The proposed API in this paper aims to secure images by utilizing a well-known Data Encryption Standard (DES) algorithm. The National Institute of Standards and Technology adopted IBM's Data Encryption Standard (DES) algorithm, a symmetric-key block cypher, in the early 1970s (NIST). Using 48-bit keys, the method transforms plain text, which is provided in 64-bit blocks, into ciphertext. Because DES can only convert plain text into secure cypher text, the Image encryption API first converted the image into a binary array so that the DES algorithm could be applied to the image in binary form. Using the API, each image to be transferred is first encrypted by selecting an encryption key. The image can only be opened by the receiver using the same key and API that the sender used during encryption. The primary need for this API or model is to ensure communication security and privacy.
Human action recognition has important applications in the motion detection, timely tracking, and comprehensive behavior analysis. It is a direction of the great scientific research significance in the field of computer vision. This study focuses on analyzing the real-time swimming posture image correction framework based on novel visual action recognition algorithm. To begin with, the body image digitization process is applied. First, this study divides the video into frames, extract the coordinates of the human body frame through the pre-trained CNN and get the annotated images for digitalization, then the marked frontal posture data is preprocessed through the translation and normalization operations. After this, the proposed OF-STH network is applied to finalize the swimming posture image correction framework. In the experiment section, the different angles are tested and the pseudo color image recognition and labeling is also tested.
A facial recognition system can be developed using a machine learning approach that involves data collection, preprocessing, feature extraction, model training, evaluation and testing, and deployment. The system can be trained on a large dataset of facial images using techniques such as PCA, LBP, or CNNs for feature extraction and SVM, Random Forest, or Neural Networks for model training. The performance of the system can be evaluated using a test set, and the system can be deployed in real-world scenarios. However, it is crucial to consider the ethical and privacy implications of facial recognition technology and implement appropriate safeguards to prevent misuse. The Eigenface, Fisherface, and LBPH (Local Binary Patterns Histogram) algorithms are three popular techniques for face recognition in the OpenCV library. This work evaluates the performance of each algorithm on a specific dataset to determine which algorithm is the most appropriate for this application.
Smart Ambulance and Patient Health Monitoring is a system designed to enhance the quality of medical care during patient transport. it is a cutting-edge technology that integrates healthcare with transportation It aims to improve the efficiency of emergency medical services. This work is an effort to address a critical issue in modern healthcare delivery. It consists of three major sections. First, sensors would be used to detect the patient's vitals; second, data would be sent to a cloud storage service; and third, the discovered data would be made available for remote viewing via a Java GUI. The ambulance is equipped with a real-time communication system that connects it with the database, enabling healthcare professionals to remotely monitor and advise on patient care in Java GUI. The patient's vital signs (heart rate, respiration rate, and temperature) are tracked in real time by sensors and wireless communication devices in the patient health surveillance system. This information is transmitted to the Java GUI including ambulance safety parameters like Fire sensor, IR sensor, GPS tracking and Gas sensor, enabling healthcare professionals to make informed decisions regarding patient care, and to enhance the ambulance's ability to reach the hospital safely. The system aims to improve patient outcomes by providing timely and accurate medical interventions during transport and may reduce the time between diagnosis and treatment.
