This paper presents a proof of concept for realtime medical signal processing using audio as the transmission medium. Vital sign such ECG and PPG are used as an example, while the approach can be applied to any biomedical signal that changes slowly or moderately over time. The system architecture includes a transmitter and a receiver. The transmitter senses the medical signal, encodes it, and converts it into an audio waveform, which is transmitted via wire, air, or fluid. The receiver captures the sound, converts it back into an electrical and digital signal, decodes it, and processes it further. Low-cost components such as standard piezo speakers and microphones are used for signal conversion, while the processing algorithms run on affordable hardware or any mobile device. For airborne transmission, the system operates in the 16-22 kHz frequency range, which is typically inaudible or barely audible to humans. The system is capable of transmitting analog or digital signals, in single or multi-channel modes. The paper describes HeatlthTalk methodology on the system level with emphasis to software decoding on receiver side, implemented in open- source, platform-independent technologies such as JavaScript + CSS + HTML. The HealthTalk is low-cost, energy-efficient, easy to implement, eco friendly and robust against noise. It can operate without a persistent connection, making it suitable for "plugout" or offline environments. HealthTalk offers a potential alternative to wireless communication systems such as RF, Bluetooth (BT), and Bluetooth Low Energy (BLE) in appropriate applications.
The document discusses the development and evaluation of machine learning models for detecting credit card fraud in an imbalanced dataset, specifically using XGBoost and Random Cut Forest (RCF) algorithms. Credit card companies must identify fraudulent transactions to avoid charging customers for unauthorized purchases, using a dataset with 0.173% fraudulent examples out of 284,807 total examples. The initial solution used Amazon SageMaker and the RCF algorithm for unsupervised anomaly detection, which identified anomalies based on their characteristics and provided anomaly scores for test cases. The XGBoost supervised learning method with weighting schemes outperformed the RCF approach, with further enhancements possible using techniques like SMOTE and Hyperparameter Optimization (HPO). Improvements to the XGBoost algorithm involved transforming the binary problem into a multi-class problem using the Majority Oversampling Technique (MOTE), leading to better performance metrics. The evaluation metrics for the XGBoost MOTE model included balanced accuracy, Cohen's Kappa, F1 score, logistic loss functions, balanced logistic loss functions, and AUC, showing significant improvements over other models. The XGBoost MOTE model was compared with other models like XGBoost with built-in weighting, XGBoost with SMOTE, and RCF, showing superior performance in most metrics.
The deployment of deep learning models on edge devices is characterized by significant challenges due to limitations in computational power, memory and energy efficiency. In particular, quantization has evolved into a leading technique for reducing the size of the model and improving the speed of inference, while still mitigating the accuracy drop. This study analyzes the performance of quantization as an optimization method and explores the feasibility of training ML models directly on IoT devices. It focuses on the impact of quantization on two neural network models: LSTM (Long Short-Term Memory) and FNN (Feedforward Neural Network) using three schemes (tf, Float32, Int8). In a comparative analysis, throughput, memory consumption, latency and symmetric mean absolute percentage error (SMAPE) were used. The results show that Int8 quantization substantially reduces memory usage and improves latency and throughput, especially for the FNN model, which achieved 99,000 samples per second at a latency of just 0.01 ms. Despite its stable performance characteristics, the Int8 scheme provided benefits to the LSTM model. Deep learning models achieve better efficiency in real-time applications through Int8 quantization while preserving accuracy with minimal loss and gaining substantial performance benefits.
In this paper, we propose a hybrid system suitable for time-frequency (TF) signal analysis. This system applies the S-method for TF signal analysis and introduces an architecture that balances two traditional extremes. A single-clock-cycle (SCI) design delivers the fastest execution but demands prohibitively high hardware complexity, whereas a multiple-clock-cycle (MCI) design reduces hardware requirements at the cost of slower execution. The proposed solution blends the strengths of both approaches, trimming hardware complexity while keeping execution time short enough for practical use. Trade-offs and comparisons of the proposed design regarding the critical design performance and the competitive designs are also provided.
Information security helps in protecting the confidentiality, authenticity and integrity of user data. In today’s digital era, steganography within the field of information security has become an important data hiding technique, in this way addressing the need for secure data transmission. It enhances information security by providing an additional layer of confidentiality, allowing the sensitive data to be hidden within different media to protect it from detection or access. The type of media that steganography uses to protect the data are videos, audios and images, which is the media carrier that we are going to use. Meanwhile, image classification helps large image datasets to be processed efficiently with machine learning and deep learning models becoming essential in this area. We are going to evaluate the impact of steganography on the performance of machine learning models like Support Vector Machine, K-Nearest Neighbors, Logistic Regression and Random Forests, but also deep learning models like Residual Network with 50 layers and 18 layers, Vision Transformer and Visual Geometry Group with 16 layers.
The deployment of secure and resilient FithGeneration (5G) core networks in industrial environments necessitates adopting methodologies that integrate security seamlessly into the entire development lifecycle. This paper presents a Development, Security, and Operations-based (DevSecOps) Continuous Integration/Continuous Deployment (CI/CD) pipeline designed to enhance the security of distributed 5G core networks. The proposed pipeline systematically analyzes Docker images and Kubernetes templates, detecting and mitigating vulnerabilities early in the deployment process. Automated tools such as Trivy, Checkov, Kubelinter, and Kube-bench are employed to significantly reduce security risks associated with outdated dependencies, misconfigurations, and network vulnerabilities. Additionally, integrating Lamassu IoT, a specialized PKI, enables automated digital certificate management. Results demonstrate substantial improvements in security practices, showing a marked reduction in vulnerabilities and enhanced compliance with industry standards.
The production of optical lenses demands stringent quality control to ensure compliance with international standards, particularly ISO 10110-7 for surface defects. Traditional manual inspection methods are labor-intensive and subjective, necessitating automated solutions. This study presents a costeffective, microscope-based imaging system for defect analysis in plano-parallel lenses, integrating a high-resolution Euromex camera, Schoelly Flexilux LED illumination, and ImageFocus Alpha software. The system achieves a resolution of 0.8 mu m/pixel, enabling detection of scratches =1 mu m and pits =50 mu m(2), while adhering to the 5% total defect area limit. Calibration with NISTtraceable standards ensures traceability, and automated workflows reduce subjectivity in defect assessment. Experimental results demonstrate successful defect identification, including chips and scratches, validated against reference etalons. The proposed system lays the groundwork for future AI-driven quality control, offering scalability for industrial applications.
A Constraint Satisfaction Problem (CSP) is a framework used to describe and solve problems where decisions need to be made under certain rules or limitations. A CSP, at its core, involves finding values for a set of variables while ensuring that all the given constraints are satisfied. A CSP consist basically of 3 main components: 1. Variables, or elements of the problems that need a decision or to assign values to, such as tasks, locations or colors. 2. The domains, which is a set of possible values for each variable, 3. the Constraints which are rules which define what combinations of values are allowed.A comparative analysis of decomposition techniques used for solving CSPs is the goal of this paper.
This research presents the design, development, and testing of an autonomous drone system for medical delivery in regions affected by natural calamities or where human access is restricted. The system leverages a waypoint navigation algorithm to enable precise and reliable delivery of medical supplies. Equipped with a GPS(Global Positioning System) module, the drone calculates destinations using the conversion of global positioning coordinates into NEU (North, East, Up) vectors, ensuring accurate navigation along predefined routes. This study highlights the potential of UAV(Unmanned Aerial Vehicle) technology in healthcare logistics, offering a rapid and adaptable solution for delivering critical supplies during emergencies. The findings underscore the transformative role of autonomous drones in bridging gaps in healthcare accessibility, particularly in disaster-stricken or remote regions.
The article discusses the use of different E-services for teaching future engineers English at technical universities. Different online boards, platforms for creating interactive tasks, games, etc. are considered. The results of the survey conducted in National Research University "Moscow Power Engineering Institute" asking lecturers and students’ attitude to using interactive tasks are presented in the article.
This paper presents the development of an experimental system based on the LabVIEW platform for measuring the electrical and thermal characteristics of aluminum conductor steel-reinforced (ACSR) conductors. The system is designed to enable realistic simulation of both static and dynamic loading conditions, including the influence of current, temperature, wind, and solar radiation, within the framework of the Dynamic Line Rating (DLR) concept. The key components of the experimental setup are described: a high-current source, measurement equipment (current probes, temperature sensors, signal integrators, oscilloscope, and analyzer), as well as an integrated data acquisition and processing system based on the NI CompactDAQ architecture. Through testing of an actual conductor sample, it has been demonstrated that the developed system is suitable for accurate analysis of temperature-dependent AC resistance, making it applicable in laboratories for testing and modeling transmission lines under real operating conditions.
This paper focuses on the modeling, simulation and optimization of a six-degree-of-freedom (6DOF) robotic arm designed for industrial applications such as assembly, material handling and precision machiing. The study utilizes the Denavit-Hartenberg (DH) parameterization for kinematic modeling. The robot, a Dobot CR5 is modeled in SolidWorks and integrated into MATLAB/Simulink for trajectory optimization using PID controllers. Simulation results demonstrate significant improvements in trajectory accuracy and motion smoothness, validating the proposed optimization approach. This work contributes to advanced control strategies for industrial robotic systems.
Accurate motion tracking is critical in robotics, augmented reality, and assistive technologies. However, low-cost Inertial Measurement Units (IMUs) often suffer from drift and noise. While standard Kalman filters are commonly used to address these issues, their fixed noise parameters limit their effectiveness across varying gesture dynamics. This paper presents a novel gesture-context adaptive Kalman filter that uses a Random Forest classifier for gesture recognition and a neural network to adjust the measurement noise covariance dynamically. Results show significant improvements over standard Kalman filtering, reducing pitch Mean Absolute Error (MAE) from 4.22 degrees to 1.70 degrees and increasing Signal- to-Noise Ratio (SNR) from 2.44 dB to 5.16 dB, with similar gains for roll. This approach advances adaptive tracking performance in wearable, human-interactive systems.
The creation of an autonomous robotic arm with a computer vision system for object detection and classification based on color and position is the main goal of this project. High precision object detection and manipulation are made possible by the combination of technologies like OpenCV and electrical microcontrollers [1]. Algorithms for image processing help with color analysis and coordinate determination enabling precise hardware-software communication. To increase productivity and decrease human interaction, the project has important applications in material classification, industrial automation, and other logistical activities. This system offers a practical solution for autonomous object identification and manipulation by using a methodical approach to mechanical design and programming. Algorithms for image processing help with color analysis and coordinate determination, enabling precise hardware-software communication. To increase productivity and decrease human interaction, the project has important applications in material classification, industrial automation, and other logistical activities. This system offers a practical solution for autonomous object identification and manipulation by using a methodical approach to mechanical design and programming.
The rapid advancement of Large Language Models (LLMs) has unlocked new opportunities for time-sensitive intelligent systems, including drone-based image analysis. However, integrating LLMs into drones presents significant challenges, particularly in balancing computational performance, latency, and energy consumption. This study investigates and compares two processing architectures, local processing and cloud processing for enabling low latency image analysis on drones equipped with LLMs. Through empirical evaluations on a forest fire detection use case, we assess various LLMs based on key performance metrics, including inference time, round-trip time, accuracy, F1 score and inference cost. Our findings highlight the trade-offs between these architectures, showing that the optimal choice depends on specific operational requirements and constraints.
This paper addresses the challenges of glucose regulation in Type 1 Diabetes (T1D) by developing an uncertain model identification framework. The approach utilizes small random variations in insulin delivery rates to identify a robust control model. The Hovorka model is adapted to incorporate uncertainty, leading to a linear time-invariant (LTI) model with multiplicative uncertainty. A robust controller is designed to ensure stability and performance under model uncertainty. The controller effectiveness is validated using the FDA-approved UVa/Padova simulator, demonstrating successful glucose regulation within the target range.
The goal of this paper is to look at how we can use AI to personalize learning platforms in a way that improves the experience for students. Rather than a one-size-fits-all approach, the idea is to use machine learning to tailor content and recommendations to each individual user. The machine learning algorithms in the proposed framework will dynamically adapt to the changes pertaining to the learning experience based on the individual user's interactions. So as you interact with the system, it gets smarter about what kinds of things you might be interested in or find helpful and the specific proposal involves building this type of personalized AI into a learning management system using Blazor components. Blazor allows content to be dynamically updated in real-time, which suits this personalized approach. We present a practical use case of the University of Tetova. By this study, we would like to underline the potential of AI-driven personalization in LMS environments and the advantages of using Blazor for real-time, modular system integration.
The evolution of artificial intelligence (AI) marks a pivotal paradigm shift in contemporary society, heralding profound changes in individual lives, societal norms, and economic landscapes. Particularly in Montenegro, the integration of AI into everyday business practices and consumer interactions unveils a plethora of advantages, encompassing enhanced operational efficiency and a heightened level of user engagement. However, alongside these benefits, AI's trajectory raises critical concerns about its future development and potential societal implications. The formidable data processing capabilities of AI pose a risk of surpassing human decision-making, potentially leading to consumer manipulation and an increased reliance on intelligent technologies. This intricate relationship between users and AI has implications for cognitive abilities, personal identity, and social dynamics.This paper investigates the Montenegrin enterprise landscape, presenting a mediation model that explores the balance between AI's efficiency, its appeal to users, and the preservation of individual self-identity and human skills. Central to this model is the role of the social circle as a mediator, influencing perceptions and interactions with AI. Our findings reveal that a higher degree of efficiency and appeal in AI, coupled with positive social reinforcement, can mitigate concerns about the identity erosion in an AI-dominated era.
The adoption rate of electric vehicles is continuously growing, leading to cleaner and more efficient transportation systems. There are also challenges caused by this increase, such as a higher impact on the energy grid or the need for better electrical vehicle supply equipment for facilitating the charging operation. Most car manufacturers include an in cord-control and protection device with a newly purchased electric vehicle for charging its battery. The control pilot plays a central role within such devices and represents a way to implement the communication between an electric vehicle and the charging equipment. It is used for transmitting the charging current information and the state of the charging session through an 1 kHz pulse width modulated signal. The current paper presents the design, implementation, and testing of the control pilot communication using commercially available off-the-shelf parts. Testing showed that the developed system correctly determines the states of the vehicle during charging and sends the appropriate information in this time. The used software and hardware components lead to a simplified architecture and reduce manufacturing time.
Aflatoxin contamination poses a significant risk to all nuts, including pistachios, during harvest, storage, and processing. Dietary exposure to aflatoxins can lead to severe toxic and carcinogenic effects in humans. To safeguard human and animal health, aflatoxin legislation sets maximum permissible levels for aflatoxins in food products, including pistachios. Consequently, imported pistachios undergo rigorous aflatoxin contamination testing. Traditional methods for measuring aflatoxin levels, such as High-Performance Liquid Chromatography (HPLC), HPLC with Mass Spectrometry, and Enzyme-Linked Immunosorbent Assay (ELISA), although precise, are destructive, costly, and time-consuming. This paper investigates the application of emerging technologies, including Hyperspectral Imaging, Chromatographic Test Strips, Luminescent Metal-Organic Frameworks, spectroscopic methods, machine vision, and advanced artificial intelligence models, to develop a non-intrusive, real-time system for aflatoxin detection in pistachio nuts. Additionally, it outlines a comprehensive strategy to protect public health, mitigate economic losses estimated at $932 million annually, and sustain the pistachio industry.
