Detection and prediction of the onset of seizures are among the most challenging problems in epilepsy diagnostics and treatment. Small electronic devices capable of doing that will improve the quality of life for epilepsy patients while also open new opportunities for pharmacological intervention. This paper presents a novel approach using machine learning techniques to detect seizures onset using intracranial electroencephalography (EEG) signals. The proposed approach was tested on intracranial EEG data recorded in rats with pilocarpine model of temporal lobe epilepsy. A principal component analysis was applied for feature selection before using a support vector machine for the detection of seizures. Hjorth's parameters and Daubechies discrete wavelet transform coefficients were found to be the most informative features of EEG data. We found that the support vector machine approach had a classification sensitivity of 90% and a specificity of 74% for detecting ictal episodes. Changing the epoch parameter from one to twenty-one seconds results in changing the redistribution of principal components" values to 10% but does not affect the classification result. Support vector machines are accessible and convenient methods for classification that have achieved promising classification quality, and are rather lightweight compared to other machine learning methods. So we suggest their future use in mobile devices for early epileptic seizure and preictal episodes detection.

With rapid digitization and digitalization, drawing a fine line between the digital and the physical world has become nearly impossible. It has become essential more than ever to integrate all spheres of life into a single Digital Thread to address pressing challenges of modern society – accessible and inclusive healthcare in terms of equality, and equity. The techno-social advancements and mutual acceptance have enabled the infusion of digital models to simulate social settings with minimum resource utilization to make effective decisions. However, there is a significant gap in feeding back the models with appropriate changes in real-time. In other words, an active behavioral modeling of modern society is lacking, that influences community healthcare as a “whole”. By creating virtual replicas of physical systems, digital twins can enable real-time monitoring, simulation, and optimization of urban dynamics. This paper explores the potential of digital twins to promote inclusive healthcare for evolving smart cities. We argue that digital twins can be used to:Identify and address disparities in access to healthcare servicesFacilitate community participationSimulate the impact of urban policies and interventions on different groups of peopleAid policy making bodies for better access to healthcareThis paper proposes several proposals for stitching the actual and virtual societies using digital twins. Several discussed concepts within this framework envision an active, integrated, and synchronized community aware of data privacy and security. The proposal also provides high- level step-wise transitions that will enable this transformation.

The advent of 6G non-terrestrial networks (NTNs) offers unprecedented opportunities for continuous Earth observation (EO), enabling enhanced global monitoring capabilities. However, NTNs face significant challenges, including high latency, signal disruptions, and bandwidth constraints, which hinder their ability to efficiently transmit real-time data during monitoring and connectivity provision. We propose an integrated framework that leverages multi-source semantic communication (SC) and intelligent denoising to address these limitations. By transmitting essential meaning rather than raw data, the SC framework reduces bandwidth requirements and enhances resilience to noise, optimizing network performance in NTNs. The proposed multisource satellite fusion and denoising framework within the 6G NTN framework enables the integration of diverse satellite data sources and improves data quality in adverse monitoring and communication conditions. We evaluate the proposed framework through important simulations demonstrating significant improvements in bandwidth efficiency, semantic accuracy, and robustness against channel impairments. The paper also highlights future research directions, including further integrating advanced machine learning (ML) techniques and adaptive denoising algorithms to enhance 6G NTN-based EO and satellite connectivity capabilities.

In this paper, we propose and investigate a novel fall detection scheme based on a wearable device. The device comprises of an FX1901 load cell placed in the sole of a footwear for monitoring the weight exerted by the body during locomotion and an ADXL335 tri-axial accelerometer for sensing tilt and associated angular displacement. Real world dataset from the sensors' output corresponding to the unique patterns that characterize the four basic activities of daily living were obtained and fed into multi-class learning algorithm for building an offline decision model. When a fall occurs, the weight exerted by the body on the sensor drops rapidly and for an unusually prolonged period of time. This is accompanied by a drastic variation in angular displacement relative to a reference position. The joint variations are processed via an Arduino LilyPad ATmega168V microcontroller incorporating the decision model of the learning algorithm and a request for help along with the subject's location information is generated and sent to caregivers by using a GPS/GSM module. Experimental results indicate that proposed scheme has the potential to detect fall with 100% accuracy.

Vehicular Ad-Hoc Networks (VANETs) are connected networks of vehicles and infrastructure that allow for communication between parties. VANETs are ever changing now that technology in and around vehicles is rapidly changing as well. Sensors, personal devices, and autonomously driving vehicles all pose new security threats to smart cities. Elliptic Curve Cryptography (ECC) can offer increased security and lowered computational effort in many of these scenarios. Our work will systematically review how ECC is being applied in real world scenarios and how it can be applied in the future to improve vehicle security in VANETs.

Data encryption is an essential element to a functioning network of networks, otherwise known as the internet. It is important for users to maintain security, whether it be their online assets or secure communication amongst other parties via a network. Data encryption algorithms are put in place to ensure these values, where AES and DES are the two most well known variations. The core purpose of this paper will be to determine why the AES algorithm has proven itself over the DES algorithm as a community preferred standard, mainly in reference to the field of CyberSecurity. This paper will look at the key differences between the DES and AES algorithms. This will include the cost efficiency differences in terms of computational power between the two algorithms, as well as the reason why the AES algorithm is more easily adaptable in today's society. This comparison will be made by dating back to the origins of both algorithms, as well as noting on why AES implementations are typically more accepted in CyberSecurity, as well as other real world sectors such as finance, healthcare and government. Key points will be made on why the DES algorithm has been proven to be susceptible to encryption attacks over time, and why the AES algorithm is able to overcome these faults by looking at their statistical differences in terms of block/key sizes, and the speed at which the algorithms take to encrypt a message. By utilizing real world examples that detail breach attempts on DES algorithms, a clear notion can be established on the weaknesses of the algorithm and how these cons were improved upon through the AES algorithm. This paper will also include the future of both algorithms by comparing them to newer encryption models and techniques, touching on how the currently accepted AES model can continue to be adapted to the newly expected era of technology and cryptography.

For powered prosthetic legs to be viable in everyday situations, they require an activity classification system that is not only accurate but also straightforward to understand and use. However, incorporating the numerous activity modes in real-world ambulation often requires high-dimensional feature spaces and restrictions on the leg leading each transition. This paper addresses these challenges by delegating sit/stand transitions and variable-incline walking to the mid-level controller, effectively reducing the classification space to four states with easily distinguishable features. We implement simple heuristic rules for both prosthetic-led and intact-led (i.e., ambilateral) transitions, using lower-limb kinematic features, ground contact and inclination, and environmental distance from an ultrasonic sensor. Two transfemoral amputee subjects using a powered kneeankle prosthesis demonstrated an ambilateral transition accuracy of 99.2% under both self-paced and rapid-paced/fatiguing conditions, with a 100% recovery rate due to backup logic or user-cued resets. The incline estimator enabled the prosthesis to continuously adapt between level and inclined surfaces without explicit classification. These results and an outdoor multi-terrain demonstration indicate that simple and straightforward transition logic can enable powered prosthetic legs to be used reliably across a broad array of daily activities.

This paper explores the cost analysis of vehicle recalls between 2010 and August 2024, focusing on Ford, Tesla, and Toyota. The study highlights Tesla's significantly lower recall volumes and costs compared to traditional automakers, despite rapid growth in the electric vehicle market. This research investigates the importance of integrating software-based solutions and vertical manufacturing processes to reduce recall costs. The approach involved analyzing recall data from the National Highway Traffic Safety Administration (NHTSA). The analysis is primarily based on categories such as air bags, electrical systems, and powertrain issues, considering other major concerns for each automaker and their financial implications. The results show that Tesla's reliance on software updates and fewer mechanical components results in fewer recalls, lower logistics costs, and more efficient issue resolution. In contrast, Ford and Toyota face higher recall volumes due to mechanical complexity. The study concludes that Tesla's innovative approach to recalls sets a benchmark for cost-effective recall management, offering traditional automakers a blueprint for reducing recall costs and improving operational efficiency.

Recent innovations in genetic technologies and epigenetic analysis have reshaped personalized medicine (Wu & Li, 2022). Additionally, employing encryption and data management mechanisms—such as hash/token frameworks integrated with blockchain-based approaches—offers enhanced data security, transparency, and efficient governance in the digital age (Chen et al., 2022). This paper introduces a novel perspective on combining DNA data and epigenetic profiles with hash/token technology , underscored by blockchain solutions to manage large-scale storage, authentication, and transactions of biological data. The proposed ecosystem is implemented via the Internet of Personal Health Things (IoPHT), augmented by AI-driven analytics, quantum-safe cryptography, and stringent regulatory compliance (HIPAA, MHRA, ISO, etc.). By elucidating the theoretical basis of DNA tokenization, presenting a methodological roadmap for real-world validations, and highlighting how this synergy could transform preventive healthcare and personalized treatment, the paper aspires to lay the groundwork for a fully regulated marketplace in genomic–epigenetic data exchange. Addressing pressing issues of data integrity, consent, and multidisciplinary collaboration, our framework paves the way for a new personalized healthcare industry , empowering both patients and key stakeholders (Brown et al., 2023).

Learning-based Joint Source-Channel Coding (JSCC) methods have recently achieved remarkable progress in wireless communication. Their elegant and efficient encoderdecoder architecture holds great potential to support future Semantic Communication (SC) technology. However, due to the inherent limitations of conventional autoencoder transmission frameworks, these methods are highly sensitive to changes in Physical Channel Conditions (PCC), such as variations in channel noise, bandwidth, and nonlinear interference. These sensitivities lead to significant degradation in coding performance, often falling below that of digital transmission schemes. Drawing inspiration from the success of the U-Net model in fine-grained tasks, we propose a novel JSCC learning paradigm called Primary Semantic Supplement (PSS). Specifically, during the training phase, we utilize the additive skip connections of the U-Net model as a potential Semantic Transmission Medium (STM) at the source-channel decoder to alleviate the semantic loss induced by PCC variations. In the testing phase, to align with standard JSCC systems, we truncate the STM and activate the source-channel decoder to generate supplementary semantics to compensate for missing semantic content. This approach offers several key advantages: first, it is robust to substantial fluctuations in PCC, ensuring high transmission accuracy even at extremely low signal-to-noise ratio. Furthermore, it maintains the flow direction of the latent semantics and facilitates the design of a lightweight baseline model. Finally, on benchmark image datasets (CIFAR10 and Kodak), our method achieves up to 88.24dB improvement in mean squared error compared to the existing state-of-the-art methods.

This study aims to assess accuracy in detecting patient presence or absence by using a bed sensor based on mmwave radar technology above the patient bed. Methods: Patients and healthy volunteers were observed during their presence or absence in a bed in hospital and home location. These observations were compared with data coming from bed sensor monitoring patient presence using tina.care® bed sensor ASWA. Results: A total of 53 different observations were performed during the study period and the bed sensor reached accuracy of 94%, precision of 90%, sensitivity of 99% and specificity of 89% to detect presence or absence of patients in a bed. Conclusions: The sensor demonstrated strong performance in detecting patient presence in bed, with reasonable specificity and low false negatives. Further research should assess bed-exit and bed-entry events, system's accuracy in a larger cohort, its impact on patient care, and the precision of vital health parameters measured by the sensor in order to compare it with similar studies.

Electromagnetics, radio frequency (RF), microwave, and antenna engineering are generally viewed as highly theoretical subjects, with a notable decline in student interest. This paper introduces a student-centred laboratory exercise for electromagnetic courses, developed as part of an IEEE Antennas & Propagation Society Educational Initiatives Programme. The proposed activity is based on transmission line analysis, measurements, and modelling. Students are required to model a set of 6 GHz transmission lines and passive networks using electromagnetic, closed-form, and their own equivalent circuit models. In addition, the modelling is complemented by measurements of varying quality and accuracy, allowing students to develop subject-specific critical analysis skills. Supporting instructors with instrument choice, we compare the measured response of the test lines using different inexpensive VNAs, evaluating their suitability to the different exercises. Thus, presenting for the first time guidance on which electromagnetic concepts could be illustrated using the different NanoVNAs. The exercises can be tailored for in-person or remote delivery. Implemented in a final-year undergraduate and Masters-level course, student feedback showed over 20% improvement in student satisfaction across all metrics surveyed.

Human digital twin (HDT) is envisioned as a system interconnecting physical twins (PTs) in the real world with virtual twins (VTs) in the digital world, enabling advanced humancentric applications. Unlike optimizing the quality of experience (QoE) of users in single-modal signal transmission for conventional services, users' QoE in multi-modal signal transmission required by HDT is difficult to guarantee. To tackle this, we study an optimization of QoE in multi-modal transmission, focusing on joint visual and haptic signal feedback transmissions from VT to its PT, for providing immersive interactions in HDT. To evaluate a synthesized performance of visual and haptic experiences, we design a comprehensive QoE model, taking into account video quality, continuous video quality switching rate and average haptic feedback error. Then, to maximize QoE with a guarantee on synchronization between visual and haptic signal transmissions, we dynamically optimize bandwidth allocation, bitrate and rendering mode of the video, and haptic signal's compression threshold. To this end, we propose a deep reinforcement learning based algorithm, called VisHap. Furthermore, we build an HDT multi-modal interaction platform for collecting an authentic dataset, and by using it, we conduct experiments, showing that VisHap is not only feasible but also outperforms the counterparts.

Walking on irregular terrains is a common situation in everyday life. The accurate detection of gait events is of paramount importance for characterizing and analyzing gait. While several algorithms have been proposed for gait timing estimation on flat terrains, an assessment of their performance on ecological-like terrains is still lacking. The purpose of the present study is to evaluate the performance of several gait event detection algorithms, as proposed in the literature, in the temporal segmentation of gait across different terrains. Nine healthy volunteers, each mounted with 17 tri-axial inertial sensors, walked on 12 different terrains with varying slopes. Gait events, identified from a marker-based optoelectronic system, were used as the reference. Nine different algorithms were applied to the data, and their performance was analyzed in terms of precision, recall, F1-score, and detection error. The performance scores of the different algorithms were compared across conditions. In general, the results show a decline in performance when transitioning from flat to other terrains, which aligns with expectations as most algorithms are optimized for regular horizontal ground. However, one method (Paraschiv-Ionescu) showed superior performance, achieving near-perfect F1 scores (close to 1) across most conditions. This study also showcases the ability of the standardized EUROBENCH benchmarking framework to test locomotion in real-like terrain conditions.

5G mmWave offers a high-precision positioning solution, functioning effectively in both line-of-sight (LoS) and operable non-line-of-sight (NLoS) conditions. However, in scenarios with complete signal blockage, integrating with motionbased models becomes crucial. This integration is achieved through Bayesian-based estimators, which entail a prediction and a correction stage weighted by their respective covariance matrices. Although covariance matrices of different prediction models have been extensively studied in the literature, the measurement covariance matrix derived from 5G-based position computations remains largely unexplored. In this paper, we propose a measurement covariance matrix tuning scheme based on a data-driven Cramér-Rao lower bound CRLB model. We validate the proposed algorithm within a simple linear Kalman filter (LKF) positioning framework. The methodology was tested in a controlled simulation scenario using a real 24-minute-long vehicular trajectory in a deep-urban environment. The results demonstrate that the developed data-driven model is reliable, maintaining a standard deviation error of less than 7 cm for 95% of the time and less than 0.5 m for 100% of the time relative to the true computed CRLB. The proposed adaptive KF sustains a position error below 30 cm for 99.3% of the time.

Generative Adversarial Networks (GANs) represent a revolutionary framework in unsupervised learning, where a generator and a discriminator interact adversarially to produce and classify synthetic data. This paper restructures the GAN training dynamics through a game-theoretic lens, introducing Nash bargaining and the take-it-or-leave-it (TIOLI) mechanism to model generatordiscriminator interactions. By framing the interaction as a cooperative bargaining problem, the proposed framework emphasizes balancing utility for the discriminator and cost minimization for the generator. Theoretical results demonstrate that the framework achieves stable and efficient GAN training, aligning the objectives of both entities. This work contributes to a deeper theoretical understanding of GANs and reinforces the theoretical foundation of GANs.

This paper studies the feasibility of direct flip-chip bonding of silicon carbide (SiC) chips onto alumina substrate for high-temperature applications up to 600℃. SiC dummy chips were prepared with sputtered thin-film Ti/TaSi2/Pt metal conductive pads, while alumina substrates were made with screen-printed thick-film gold conductive pads with a film thickness of about 12 µm. Thermocompression flip-chip bonding was implemented with gold stud bumps made using a wire bonder. Two strategies were used in the flip-chip bonding and compared: bump-on-substrate (BoS), where gold stud bumps were made on the conductive pads of the alumina substrate, and bump-on-chip (BoC), where gold stud bumps were made on the thin-film conductive pads of the SiC chip. The SiC flip-chip packages showed a trend of decreasing daisy chain electric resistance after thermal aging and stabilized. Die shear tests and daisy chain electrical resistance measurements were carried out for the BoS and BoC packages before and after up to 16 days of thermal aging in lab air at 600℃. Die shear tests showed that the bonding strength of the packages increased from about 15 gram force per bump (gf/bump) before thermal aging to about 40 gf/bump after thermal aging because of the effects of thermal annealing.

Alzheimer’s disease (AD) is a prevalent neurodegenerative disorder that leads to a range of cognitive complications. Early diagnosis is essential in delaying disease progression, improving quality of life, and reducing treatment costs. Unfortunately, early diagnosis is difficult due to the subtle initial brain changes. Existing diagnostic methods suffer from issues relating to accuracy, accessibility, cost, and invasiveness. This study addresses these limitations by leveraging machine learning (ML) to analyze handwriting data as a non-invasive diagnostic tool. Gaussian Naive Bayes (Gaussian NB), GradientBoosting, Support Vector Machine (SVM), RandomForest, and AdaBoost classifier models were trained and validated using Stratified K-fold cross validation. Models were trained on the DARWIN (Diagnosis AlzheimeR WIth haNdwriting) dataset. Techniques such as hyperparameter tuning, feature selection, and strong regularization were incorporated to limit overfitting and ensure the model’s ability to generalize. Performance metrics were acquired by testing the model with a separate validation set. The Support Vector Machine was the best model, achieving an accuracy of 94.49%, sensitivity of 93.53%, and specificity of 95.49%. My results suggest that utilizing Machine Learning to process handwriting data for Alzheimer’s diagnosis is a promising approach to revolutionize AD diagnosis. Handwriting data is easily obtainable, and the model’s simple online use makes it user-friendly and accessible. These results highlight the potential of handwriting-based ML models as accessible, cost-effective, and accurate diagnostic tools for AD. The model’s performance met or even surpassed similar machine learning works. In conclusion, a machine-learning model trained on handwriting features was developed to diagnose Alzheimer’s disease.