Recently the study of layered human tissues in the presence of electromagnetic radiation has gained a lot of attention, in order to develop low-cost and efficient methods of health monitoring. In this work, we explore the electromagnetic field response of a layered air-skin-abdominal fat-muscle system both analytically and through simulation, and propose the use deep neural networks to make a non-invasive estimation of fat thickness. First, we explore the TM planewave response of the layered system with varying abdominal fat thickness, and results show that the location of resonance (minimum reflection) shifts towards the lower frequencies with increasing fat thickness. Then, we explore the TM Ez field response of a horizontal dipole and the S11 parameters of a wideband patch antenna over the same layered system. For both cases, we note that the presence of resonance is obscure; although there remains a distinct relation between the received fields magnitude and fat thickness. Taking these relations into account, we train a deep learning model to solve the inverse problem of estimating the abdominal fat thickness from those data. For both sources, the overall estimation accuracy was very high, with maximum estimation error being < 10%.

SpaceX's Starlink mission is a well known mission all around the world for its goal of providing a high-speed network with low-latency, to remote areas all around the Earth. As of November of 2023, there are 5,420 Starlink satellites in the Low Earth Orbit (LEO), which is around 200-1600 Km above the Earth. To be more precise, it orbits at exactly 550 Km above sea level, reducing the overall latency. In this study, we studied the patterns in the formation and placement of SpaceX's Starlink satellites so as to get an understanding of the reason behind it. We hypothesized that the Starlink satellites are going to be placed along the latitudinal planes of the Earth in order to maintain a precise and organized network of satellites. And since the gravitational field in the poles is more, we also hypothesized that there might not be as many satellites in the poles as there are in the equator and tropics. Supporting our hypothesis, we found out that there were not a lot of satellites near the poles, one of the reasons being gravity.

Glioblastoma is a common and fatal tumor presenting a poor survival rate. To choose the best course of treatment, patients and providers need to predict the survival rate of patients. Historically, statistical methods have helped analyze clinical features to forecast survival, while recently the same is being accomplished by applying artificial intelligence techniques. However, most of these works are limited to predicting 1-, 2-, or 10-year survivability with several of these works simulating data for balancing the dataset. Hence, there is a need for fine-grained prognosis without tempering the data. To achieve the same, we employ data from Surveillance, Epidemiology, and End Results (SEER) along with an ensemble of classification and regression models to develop a fine-grained model to predict the survival period of glioblastoma patients. The proposed framework titled 'Peshnaja' presents higher resolution in the prognosis of glioblastoma while showcasing an accuracy of 70% with an overall RMSE of 2.65. Moreover, a comparison of Peshnaja with other frameworks shows that we did not impute missing values nor employed synthetic data to force good results, thereby keeping Peshnaja true to the existing data.

Exoskeleton robots hold tremendous promise in aiding the rehabilitation of patients with lower limb motor dysfunction. However, their effectiveness relies on the real-time and accurate prediction of knee joint angles. This study introduces KINETIQA (Kinematic Integration Network with Enhanced Temporal Intelligence and Quality-driven Attention), a novel model that significantly pushes the boundaries of knee angle prediction, offering substantial enhancements in clinical practice. The core strength of KINETIQA lies in its Kinematic Integration Network (KIN), a sophisticated architecture seamlessly integrating state-of-the-art techniques such as Transformers, temporal convolutions, and multi-headed attention mechanisms. This powerful synergy equips KINETIQA to capture the intricate dynamics of knee movement, both in the blink of an eye and over longer durations, resulting in a notable 20% reduction in prediction error compared to existing models. In addition to accuracy, KINETIQA employs rigorous feature selection and validation methods to guarantee clinical-grade precision. These processes offer clinicians crucial insights for tailored rehabilitation, injury prevention, and movement analysis. Our model's pioneering design integrates Transformers and temporal convolutions to capture detailed temporal dynamics of knee movement, leading to markedly enhanced prediction compared to current frameworks. These positive findings indicate that the proposed approach holds significant promise for application in exoskeleton robot control. The source code for KINETIQA is publicly available on GitHub at https://github.com/LyesSaadSaoud/KINETIQA/

This paper presents a Python-driven graphical user interface (GUI) explicitly crafted for benchmarking the electromagnetic transient (EMT) models of inverter-based resources (IBRs) across Matlab/Simulink and PSCAD/EMTDC platforms through impedance measurement techniques. The GUI ensures a user-friendly experience, enabling seamless interactions between these platforms while offering advanced visualization of impedance measurement results and associated benchmarking errors. This paper delves into the Python-based implementation that powers these features. Through a series of case studies, the tool's efficacy in benchmarking IBR's EMT models is validated. The benchmarking methodology and corresponding GUI design are extensible to other leading EMT platforms such as DIgSILENT/PowerFactory, PSSE, EMTP-RV, EMTP-ATP, and DSATools TM /SSAT, fostering expansive and holistic research in the stability analysis of the IBR-based modern power systems.

Grid-tie voltage source converters (VSCs) can operate in three distinct modes: AC-dominant, DC-dominant, and balanced, depending on the placement of the stiff voltage sources. The varying operation modes of the VSCs traditionally demand different synchronization control techniques, posing challenges for the power systems to accommodate and coordinate the VSCs. A promising universal control technique for VSCs is the DC-link voltage synchronization control (DVSC) based on a lead compensator (LC). The LC DVSC stabilizes both the DC and AC voltages of a VSC while achieving synchronization with the AC grid. This results in a dual-port grid-forming (DGFM) characteristic for the VSC. However, there has been very limited study on the stability and synchronization controller design of the VSCs with the LC DVSC operating in various modes. To bridge this gap, the paper presents a quantitative analysis on the stability and steady-state performance of the LC DVSC in all three operation modes of the DGFM VSC. Based on the analysis, the paper provides step-by-step design guidelines for the LC DVSC. Furthermore, the paper uncovers an instability issue related to the LC DVSC when the DGFM VSC operates in the balanced mode. To tackle the instability issue, a virtual resistance control is proposed and integrated with the LC DVSC. Simulation results validate the analysis and demonstrate the effectiveness of the DGFM VSC with LC DVSC designed using the proposed guidelines in all three operation modes. Overall, the paper demonstrates the feasibility of employing the DGFM VSC with the LC DVSC for all three possible operation modes.

A document by Eric J Earley. Click on the document to view its contents.

Sensor-based remote health monitoring can be used for prompt detection of adverse health in people living with dementia in the home. Current anomaly detection approaches are challenged by noisy data, unreliable event annotation and wide variability in home settings. We hypothesized that a downturn in health would present as a discernible shift in spatiotemporal patterns, which can be identified by monitoring the temporal evolution of the household movement graph. We present a lightweight contrastive learning approach to detect adverse events using home activity changes, along with household-personalized alerting thresholds based on the clinician-set target alert rate. Our self-supervised Graph Barlow Twins model with aggregation-based node feature masking is used to generate daily activity representations in participant households taken from a real-world dataset collected by the UK Dementia Research Institute. Daily graph differences represent the anomaly score, which are compared to the householdpersonalized threshold, and alerts raised to the clinical monitoring team. Attention weights from the graph encoder support explainability and help focus on the source of anomaly. Our model outperforms state-of-the-art temporal graph algorithms in detecting agitation and fall events for three distinct patient cohorts, with 81% average recall and 88% generalizability at a target alert rate of 7%. To the best of our knowledge, we offer the first use case of negative sample-free graph contrastive learning for anomaly detection in a healthcare setting that is domain-agnostic and can be applied to wider IoT settings.  

Encryption (naïve/selective) is recommended to secure the recorded visual content; however, intruders can still manipulate encrypted data. Visually, tampering attacks on encrypted video pixels in selectively encrypted videos are difficult to identify. Thus, this paper presents a tampering detection system that performs vulnerability analysis for Regions-of-Interest (ROI) in encrypted videos. To detect the tampering attacks, we explored the pixels' intensities and proposed a new TampDetect algorithm. The TampDetect applies the Hue, Saturation, and Value (HSV) colour model to detect the encrypted areas in the video. These encrypted pixels are then segmented and selected from the non-encrypted pixels using a minimum and maximum HSV value and a global threshold. The mean intensity of the encrypted pixels is thus calculated and stored. The integrity of the video frame is then validated by comparing the stored mean intensity with the newly calculated mean intensity to validate tampering/attack. The experiments were conducted on an Intel NUC, and its low computational cost demonstrates the lightweight nature of the proposed TampDetect algorithm for detecting tampering in ROI encrypted videos. The developed dataset for experiments, i.e., original, ROI encrypted, and tampered videos (encrypted/decrypted) is made available on kaggle-repository for future researchers.

Inverter-based low-voltage microgrids face the issue of imbalanced reactive power sharing among parallel inverter units due to predominantly resistive feeders. This issue can be addressed using large droop coefficients in primary power control loops. However, the large droop coefficients can lead to increased power oscillations and lightly damped modes in the system. To address this issue, this paper proposes an adaptive droop control strategy. This approach adapts the reactive droop coefficient in real-time, ensuring the best possible reactive power sharing while maintaining sufficiently high damping of the inverter oscillatory modes. The paper first derives a mathematical relationship between the damping-ratio of the associated inverter mode and droop coefficients, showing their monotonically inverse dependence. Based on the relationship, an adaptive strategy is formulated that continuously updates the reactive droop coefficient based on the estimated damping-ratio of the critical system modes. The damping-ratio is derived by perturbing the reactive power droop coefficient and processing the inverter's active power response using the Matrix-Pencil mode estimation algorithm. The proposed adaptive droop control offers three key advantages: (i) enhanced reactive power sharing, (ii) improved microgrid stability through adaptive droop coefficient adjustment, enabling inverters to act as stability reservoirs, and (iii) applicability to grids with unknown R/X ratios.

A document by James Greig. Click on the document to view its contents.

The traditional approach to designing filtering antennas involves conducting numerous simulations, which consumes a significant amount of time and computational resources, demanding a high level of expertise. This paper presents a novel approach to designing filtering antennas using the Grey Wolf Optimizer (GWO). Firstly, a structure with free-form metal patterns is constructed to fulfill the filtering function requirements. Next, the GWO is employed to optimize the parameters of the free-form metal pattern. Finally, the optimized metal pattern is integrated into the final antenna structure. The key to this approach lies in the fitness function enabling the GWO to evaluate the filtering response automatically. This method streamlines and automates the development of filtering antennas, making it possible to design them without prior knowledge of filtering principles. To validate our approach, we applied it to design a filtering magneto-electric dipole (ME-dipole) antenna, demonstrating its effectiveness and reliability.

Sea ice property retrieval from synthetic aperture radar (SAR) at high resolution has one central problem: there are no comprehensive ground truth data available. One candidate instrument to offer such ground truth is the ICESat-2 altimeter, which has excellent vertical resolution. Due to the ice motion however, both instruments have to be acquiring within a short time difference for the measurements to still be correlated. Because the ICESat-2 altimeter uses a laser at optical wavelengths, cloud free conditions are also needed. As a result, valuable coincident measurements are sparse. Deep learning methods, that have shown promising results for ice classification in the past, particularly suffer from sparse training data. In this paper we propose an architecture agnostic transfer learning and autoencoding approach to overcome the problem of data sparsity. A major component of this approach is using local incidence angle dependencies as a proxy for ice classes. Thus, we can do a majority of the deep learning model training using self-supervised training on data without labels. Using this technique, altimetry derived ice development can be meaningfully extrapolated to Sentinel-1 SAR acquisitions.

Artificial Intelligence-Generated Content (AIGC)- related network services, especially image generation-based services, have garnered notable attention due to their ability to cater to diverse user preferences, which significantly impacts the subjective Quality of Experience (QoE). Specifically, different users can perceive the same semantically informed image quite differently, leading to varying levels of satisfaction. To address this challenge and maximize network users’ subjective QoE, we introduce a novel interactive artificial intelligence (IAI) approach using Reinforcement Learning With Large Language Models Interaction (RLLI). RLLI leverages Large Language Model (LLM)-empowered generative agents to simulate user interactions, thereby providing real-time feedback on QoE that encapsulates a range of user personalities. This feedback is instrumental in facilitating the selection of the most suitable AIGC network service provider for each user, ensuring an optimized, personalized experience.

A document by Shuo Li. Click on the document to view its contents.

Feedback control plays a crucial role in improving system accuracy and stability for a variety of scientific and engineering applications. Here, we theoretically and experimentally investigate the implementation of feedback control in microwave photonic (MWP) transversal filter systems based on optical microcomb sources, which offer advantages in achieving highly reconfigurable processing functions without requiring changes to hardware. We propose four different feedback control methods including (1) one-stage spectral power reshaping, (2) one-stage impulse response reshaping, (3) twostage spectral power reshaping, and (4) two-stage synergic spectral power reshaping and impulse response reshaping. We experimentally implement these feedback control methods and compare their performance. The results show that the feedback control can significantly improve not only the accuracy of comb line shaping as well as temporal signal processing and spectral filtering, but also the system's long-term stability. Finally, we discuss the current limitations and future prospects for optimizing feedback control in microcomb-based MWP transversal filter systems implemented by both discrete components and integrated chips. Our results provide a comprehensive guide for the implementation of feedback control in microcomb-based MWP filter systems in order to improve their performance for practical applications.

A document by Onurcan Kaya. Click on the document to view its contents.

Predicting signal blockages in millimetre waves (mmWave) and terahertz (THz) networks is a challenging task that requires anticipating environmental changes. One promising solution is to use multi-modal data, such as vision and wireless inputs, and deep learning. However, combining these data sources can lead to higher communication costs, inefficient bandwidth usage, and undesirable latency, making it challenging. This paper proposes a semantic aware federated blockage prediction (SFBP) framework for a vision-aided next-generation wireless network. This framework uses computer vision techniques to extract semantic information from images and performs distributed on-device learning to enhance blockage prediction. Federated learning enables collaborative model training without exposing private data. Our proposed framework achieves 97.5% accuracy in predicting signal blockages, which is very close to the performance of centralised training. By using semantic information to train models, SFBP reduces communication costs by 88.75% and 57.87% compared to centralised learning and federated learning without semantics, respectively. On-device inference further reduces latency by 23% and 18% compared to centralised and federated learning without semantics, respectively.