The examination of hyperspectral data has become a powerful tool in the field of document image processing, offering unprecedented levels of information and insights across various applications. Traditionally, tasks like document forgery detection, ink age estimation, and text extraction from degraded or damaged documents have relied on RGB images. However, as the need for more precise tasks, such as multiclass signature segmentation, arises, there is a demand for a richer source of information to avoid data loss. This paper introduces a new method for segmenting signatures with class overlap in hyperspectral document images. Unlike conventional RGB approaches, our proposed technique is specifically designed for hyperspectral data. To validate the effectiveness of our methodology, we rigorously test the handwritten signatures segmented using our approach against ground truth images. The results confirm that our method is not only effective but also efficient and precise in the challenging task of segmenting signatures with class overlap in hyperspectral document images. This breakthrough significantly enhances the capabilities of hyperspectral data analysis in the domain of document image processing.

Along with the increase in the spread of smartphones, the use of smartphones while walking has increased, and the resulting accidents has also increased. There are signs and exclusive roads implemented to prevent accidents caused by smartphone use while walking, but these measures are limited to specific areas. So, this paper creates a system to assist pedestrians without limitations to specific locations, utilizing the LiDAR sensor of the iOS Pro model. LiDAR provides accurate height information for objects in the frontal direction compared to a camera. However, the LiDAR detection range for the front varies depending on the angle of the smartphone, and individuals may use the smartphone at different angles. To address this, the angle of the smartphone is measured using an acceleration sensor. But the angle error caused by the change in acceleration during walking. So this error is corrected using the firstorder low-pass filter, and the angle is 4.13% more accurate than before the correction. Three LiDAR ranges are set according to the angle, and the height change is detected using the first-order low-pass filter. Finally, a pedestrian assistance system is created to alert pedestrians to danger if the height is more than 10cm before.

In many futuristic communication systems, integrated sensing and communication (ISAC) is emerging as a crucial use case, with wide applicability in vehicle-to-everything (V2X) scenarios. As orthogonal time frequency space (OTFS) is being investigated as a possible waveform for high-velocity scenarios, in this work we address the challenging fractional delay-Doppler (DD) bin shifts estimation by considering an OTFS-based multistatic ISAC system. We make the problem pilot agnostic and propose a novel two-dimensional delay-Doppler (2D-DD) virtual array incorporating the bin shifts as phase information. To jointly estimate the fractional DD bin shifts, two algorithms, DD-OMP and DD-ESPRIT, are proposed using such a 2D-DD virtual array, based on orthogonal matching pursuit (OMP) algorithm and estimation of signal parameters using rota-tional invariance (ESPRIT), respectively. The proposed algorithms are further analyzed to determine: a) maximum number of targets that can be identified using DD-OMP and DD-ESPRIT; b) Cramer-Rao lower bound (CRLB); and c) computational complexity. Simulation results underscore the effectiveness of the proposed algorithms to achieve improved target parameter estimation and communication link performance. Being super-resolution schemes, this improvement in performance is reflected even at lower bandwidths and is independent of the pilot structure, thus making them attractive for ISAC systems.

The aim of this study is to develop a Washington, D.C. house price prediction model based on prior transactions in a very interesting and dynamic market. Therefore, we develop predictive models using the richness of the historical housing dataset concerning geographic location: neighborhood and quadrants, property size, number of rooms and other amenities offered, and factors such as distance from access to public transport and schools. K-means clustering was then used to address the class imbalance by defining meaningful subgroups, and thus doing random oversampling in low-represented categories for balancing. The machine learning models that have been used include Random Forest and Gradient Boosting, whose performances were then compared based on RMSE and R 2 metrics. The ensemble methods, in turn, have outperformed the more straightforward models in predictive accuracy, thus proving to be effective in capturing the complexities of housing price prediction. The comparison of machine learning approaches thus forms a valuable foundation for future studies and practical applications in real estate.analytics.

This case study examines the integration of Lean Supply Chain Management (LSCM) and Industry 4.0 technologies in European manufacturing companies. The study aims to understand how digital technologies can enhance lean practices and improve supply chain performance. Key digital tools implemented include IoT sensors, data analytics, and cloud computing, which were used to monitor and optimize supply chain operations. Employee training and real-time monitoring systems were also critical components of the implementation process.The results indicate significant cost savings due to improved operational efficiency and reduced waste. Additionally, the integration led to environmental benefits, such as a reduction in carbon emissions and waste. Enhanced visibility and control over supply chain operations improved load stability and reduced product damage.The discussion highlights the practical benefits of integrating LSCM and Industry 4.0 technologies, including enhanced efficiency, environmental impact, and operational control. The study also provides insights into strategic integration, continuous improvement, and operational synergy. Current market trends indicate a synergistic effect between LSCM and Industry 4.0, aligning with the growing emphasis on green and sustainable processes.This case study offers valuable insights for industry practitioners and policymakers on the practical benefits and challenges of integrating LSCM and Industry 4.0 technologies, supporting the feasibility and proportional benefits of this integration for optimizing supply chain performance.

This paper considers the problem of tracking a timevarying number of non-cooperative emitters based on singlechannel passive sensors, which is traditionally carried out via sequential processing, i.e., detection, measurement extraction and tracking. However, in such a framework, tracking performance is severely affected by the pre-processing steps especially in low signal-to-noise ratio (SNR) scenarios. To remedy this drawback, in this paper we develop an algorithm that directly performs tracking based on the received signals. The states of multiple emitters are represented by a labeled random finite set (LRFS), which is further modeled by a generalized labeled multi-Bernoulli (GLMB) density. A new likelihood model, which inherently exploits time difference of arrival (TDOA) of signals coming from different sensors, is proposed. Then, a new GLMB filter is derived to iteratively compute the posterior of multiple emitters. Moreover, an efficient implementation strategy is devised for the proposed algorithm and its performance is assessed via simulations.

Data availability can limit machine learning model generalizability, while privacy concerns arise when sharing data for collaborative learning. Additionally, limited device capabilities and increased communication power can hinder real-world deployment. Differential-Private-Federated-Learning (DPFL) addresses these challenges by sharing encrypted gradients to enhance privacy and minimize data vulnerability while supporting collaborative model training in privacy-sensitive environments. The integration of DPFL with low-power FP-GAs is beneficial in overcoming edge-device limitations. FP-GAs' parallel processing, reconfigurability, quantization support, and hardware-software co-design, enhance DPFL's efficiency and adaptability. This combined approach optimizes training, ensuring secure and efficient collaboration across applications. Importantly, as ML algorithms evolve, FPGA implementations promote sustainability by reducing energy consumption, extending the hardware lifespan, and reducing electronic waste. Moreover, this framework can be used for emulation on FPGAs to explore and develop novel ASIC architectures for DPFL machine learning accelerators. We propose a novel open-source framework-"DPFL-FPGA-Accel" for carrying out design-space exploration of FPGA-based DPFL, source code is made available on https://github.com/shakeelakram00/DPFL-FPGA-Accel-Framework.git for wider adoption. To the best of our knowledge, this marks the first implementation of DPFL with FPGAs that enables users to design optimized FPGA accelerators for ML tasks implementing privacy-preserving federated learning with adjustable epsilon values. We illustrate the application of this framework in processing healthcare data, designing a DPFLenabled FPGA system for cardiac arrhythmia classification within a typical hospital setting, where electrocardiograms are not shared among hospitals. The FPGA-Accel achieves significant inference speedups, providing up to 605.2x and 107.2x faster performance compared to Arm Cortex-A53 and Intel Core i7 processors, respectively.

Multiple mobile robot systems (MRSs) of different heterogeneity degrees, scales, and interaction levels have been deployed in various application domains. Multi-robot task allocation (MRTA) and motion planning (MRMP) are two key enabling research areas for such successful deployments. While significant progress has been made in both areas, recently, multi-robot integrated task and motion planning (MR-TAMP) is gaining attention from both researchers and industrial practitioners. The main objective is to obtain joint, coupled, and informed plans in both task and motion levels simultaneously to meet certain performance measures. Motivated by both theoretical significance and practical needs, this paper provides the community with a comprehensive survey of this emerging field. Firstly, a complete problem formulation is provided. Then a revised method-centric and a novel method-centric taxonomy are proposed. Based on two taxonomies, major MR-TAMP problems and approaches are surveyed and discussed. Finally, new research opportunities and future challenges are addressed. Note to Practitioners-Mobile robots are key physical equipment in both industries and services, i.e., manufacturing, logistics and warehouse automation, transportation, healthcare, search and rescue, construction, surveillance, and planetary exploration. To realize task-oriented objectives with long-term autonomy, coupled, joint, and informed decision-making at both task and motion levels is essential. This paper provides a comprehensive review of integrated task and motion planning for multi-robot systems. It will help industrial practitioners analyze emerging problems, find solutions, make reasonable decisions, and finally, improve the overall system efficiency.

The deployment of fifth-generation (5G) networks across various industry verticals is poised to transform communication and data exchange, promising unparalleled speed and capacity. However, the security concerns related to the widespread adoption of 5G, particularly in mission-critical sectors, present significant challenges. This article investigates the potential of a Zero Trust (ZT) security philosophy as a viable countermeasure to these concerns. It delves into the practicalities of implementing ZT principles within 5G networks, with a specific focus on harnessing AI/ML technologies for proactive security measures, dynamic policy adaptations, and advanced risk assessments. Further, the article underscores the importance of developing a tailored ZT maturity model for 5G networks. Furthermore, the paper outlines key future research directions aimed at improving the ZT maturity of 5G deployments, contributing to the safe and secure integration of 5G technology in various sectors.

Rainfall prediction is a vital component of meteorological science, crucial for agriculture, disaster management, and water resource planning. This paper presents a comparative study of Hidden Markov Models (HMMs) and Bayesian Models applied to rainfall prediction. Various case studies are explored to highlight the methodologies, applications, and effectiveness of these models in different regions and scenarios. The study evaluates the strengths and weaknesses of both approaches and provides recommendations for researchers based on model accuracy, computational requirements, and practical application.

The advantage of using LLMs (Large Language Models) for query understanding is that we do not need to manually annotate data. But the problem with using LLMs for query analysis in e-commerce search is that the accuracy rate can at most reach up to 88%-90%, due to the lack of domain specific information. Meanwhile, based on mining query tags from user orders, the accuracy can reach over 95%. But the coverage rate of order-based query tags is lower than 50%, which is limited by the amount of user behavior logs, especially when the business scope is small. In this article, we propose a method for constructing training dataset for query understanding/tags, which combines the data by LLMs with the order-based data. Models trained on this dataset can achieve a coverage rate of over 95% and an accuracy rate of over 95% for query tagging.

With the increasing global demand for ultra-highspeed, low-latency communication, the development of 6G networks is inevitable. However, achieving seamless global coverage faces challenges due to complex geographical conditions and dynamic demand, making traditional terrestrial networks insufficient. Low Earth Orbit (LEO) satellites, with their low latency and high spectral efficiency, play a critical role in 6G networks. This paper focuses on LEO satellite constellation orbit design and coverage optimization, proposing a dynamic optimization strategy and AI-based framework using deep learning and reinforcement learning algorithms. First, a regional priority model is built using deep learning to identify high-demand areas based on global population distribution and terrestrial network characteristics. Then, deep reinforcement learning is applied to optimize orbit parameters, adjusting satellite deployment density to enhance coverage performance and resource utilization in priority areas.Simulation results show that this strategy significantly improves global coverage and resource efficiency, especially in high-population-density and remote areas, providing higher-quality communication services. This research offers innovative design ideas for global 6G coverage and provides a foundation for dynamic optimization and deployment strategies in future communication networks.

Federated learning (FL) is a paradigm for training deep neural network (DNN) models on a selected set of IoT/edge devices owned by clients without sharing local datasets. Devices such as smartphones, notebooks, autonomous vehicles, and so on train a specified global model using local datasets in multiple rounds orchestrated by a central FL server and share model weights with the latter. The FL server updates the global model by aggregating these weights and transferring the updated model back to clients for future training rounds. One of the key aspects of FL is that the data ownership solely rests with clients, who generate them independently. Furthermore, the quality of data belonging to the clients may differ due to hardware heterogeneity, spatio-temporal disparity, client preferences, and characteristics. As a result, the data used for model training at different devices are non-independent and identically distributed (non-IID) and suffer from quantity, label, and feature skewnesses. Data distribution and quality heterogeneities introduce biases that prevent the global model from achieving the desired convergence. Therefore, there is a need to select clients with relatively more balanced and better-quality data to ensure performance convergence. This paper proposes IQ-FL, an Image Quality-based novel client selection mechanism for FL with heterogeneous data distribution. We formulate a metric called IId-index for the FL server to rank the clients considering image quality and data imbalances and propose a scheduling algorithm to engage them in model training, ensuring participation count is restricted within a finite upper bound. This metric preserves desirable data privacy and is based on the feature extraction information of the client's DNN model and other metadata information. We have implemented IQ-FL on an open-source FL evaluation platform called FedEval and carried out an extensive empirical study. We compared our approach with three baselines on three publicly available datasets on grayscale images. Experimental results demonstrate that IQ-FL selects clients with more balanced and better quality data, resulting in better accuracy than the chosen baselines, and the process maintains client participation under a permissible limit.

A document by ye yang. Click on the document to view its contents.

A radially periodic 2-D leaky-wave antenna (LWA) with high gain at broadside, reduced side lobes, and suppressed open stopband (OSB) is presented in this communication. The antenna is low profile and is defined by a printed, double microstrip bull-eye aperture. To support polarization reconfigurability and application to full-duplex systems, dual-polarization scenarios, etc., the antenna is fed with a compact multi-slot array at the centre of the LWA and this is enabled by substrate integrated waveguide (SIW) technology. When considering differential excitation, the simulations and measurements demonstrate a maximum realized gain of more than 17 dBi at broadside at about 18.4 GHz. Also, thanks to the suppression of the OSB, persistent broadside radiation is supported from 18 GHz to 19 GHz as well as continuous and sustained directive radiation when the main beam scans through broadside, which is in agreement with leaky-wave theory. Applications include satellite connectivity, polarization diversity scenarios, radar and monopulse systems, as well as 5G/6G wireless communications.

The increasing complexity of modern machine learning models has led to significant challenges in improving their generalization and robustness during training. Stochastic Gradient Modification (SGM) offers a novel approach to address these challenges by introducing controlled random perturbations into the gradient descent process. This innovative technique allows for broader exploration of the solution space, enabling models to avoid local minima and reduce overfitting, while still maintaining training stability. Through extensive experimentation on an opensource language model, the findings demonstrate that SGM significantly enhances both training efficiency and downstream task performance, particularly on benchmark datasets such as SuperGLUE and SQuAD. The ablation study further highlights the importance of tuning perturbation magnitudes, showing that moderate levels of randomness strike the best balance between model exploration and convergence. Overall, SGM provides a promising direction for refining optimization processes in machine learning, especially in applications requiring improved generalization and adaptability across diverse linguistic tasks.

The demand for lithium-ion batteries has grown rapidly, driven by the global shift toward renewable energy, electric vehicles (EVs), and energy storage systems. In 2023, the global market for LIBs exceeded $100 billion, with a projected compound annual growth rate (CAGR) of 18% through 2030[1, 2]. Despite these economic benefits, LIB manufacturing poses environmental and economic challenges. One of the most pressing issues is the generation of waste during production, which includes defective electrodes, electrolyte residues, separator materials, and packaging.

The Study of depression and its effects on the brain is essential since this common mental health disorder affects millions. In addition to disturbing emotional and cognitive processes, depression also disrupts activity in discrete brain regions. Identifying these distortions is important for expanding the diagnosis and treatment plans. A recently introduced spiking neural network framework called NeuCube has successfully demonstrated its effectiveness in modeling dynamic brain activity using EEG signals by capturing the temporal and spatial patterns of neural activity. In this study, we introduce a new model to interpret brain region contributions in depression by integrating NeuCube architecture with a dictionary learning method. NeuCube artificially replicates neural events in multiple brain regions and those output spike trains are then combined with dictionary learning to recognize depression-related patterns. This hybrid solution allows a high level of interpretability with respect to the contribution from biology and connectivity while maintaining strong separation power for depression diagnosis. Significant contributions were observed in neuromarkers of brain regions, such as frontal and temporal lobes for eyes-closed, and eye-open states. Through the analysis of sparse codes, we reveal which brain region interactions are affected by depression, providing an understanding of the neural underpinnings behind this disorder. In addition, our model achieves an accuracy of 91% for both these conditions, stronger than traditional methods. Notably, we reach this level of performance without the use of windowing techniques emphasizing the resilience and practicality of our approach in depression diagnosis.