Functional near-infrared spectroscopy (fNIRS) has been increasingly applied in poststroke research. The accumulated evidence in this area warrants a comprehensive review systematically investigating the utility of fNIRS in poststroke rehabilitation, specifically focusing on upper limb motor recovery. The target of this systematic review was the use of fNIRS for monitoring brain function, predicting outcomes, and evaluating rehabilitative interventional responses in poststroke patients with upper limb hemiplegia. A literature search was carried out using PubMed, Web of Science, EMBASE, Medline, and IEEE Explore, to identify studies that applied fNIRS in stroke survivors. A total of 52 studies were included, with 23 cross-sectional studies, 6 longitudinal studies and 23 interventional studies. The majority of the included fNIRS studies displayed a bilateral activation pattern in patients after stroke during paretic upper limb movement. The change in hemispheric laterality, measured by oxygenated hemoglobin concentration changes (Δ[HbO]) levels in different corticomotor regions, has been found to be correlated with motor recovery following a stroke. Various rehabilitation interventions, such as exercise-based, stimulation-based, and neurofeedback techniques, improved recovery outcomes by increasing Δ[HbO] levels in the ipsilesional sensorimotor and secondary motor areas. These interventions also recruit different brain regions connected to the ipsilesional sensorimotor area, thereby strengthening their connectivity. In conclusion, outcomes derived from fNIRS demonstrate promise in monitoring brain function, predicting outcomes, and evaluating responses to interventions in patients after stroke. Future fNIRS research can be enhanced by adhering to best practice checklists, utilizing the latest experimental setup and analysis protocols, and recruiting large sample sizes.

Recently, SRAM-embedded compute-in-memory (CIM) hardware has emerged as a promising solution to mitigate von-neumann bottlenecks and has shown noteworthy improvements in energy efficiency and throughput for matrix-vector multiplication, a significant portion of neural networks. While PVT variations significantly impact traditional analog/mixedsignal (AMS) macros, the DCIM macros are more robust. This article proposes a DCIM macro that incorporates an 8-transistor SRAM bitcell capable of performing 1-bit multiplications and addressing the bit-flip issue arising from the simultaneous activation of multiple array rows. The macro also includes a 2D interleaved adder tree constructed using novel 7T-based ripple carry adder (RCA), significantly reducing the adder tree's area. The proposed 16Kb DCIM macro computes 64 parallel products in a single clock cycle. It demonstrates 2× higher energy efficiency than recent state-of-the-art works at 65nm CMOS. The macro is validated at 250MHz and achieves the classification accuracy of 98.7%, 98.8% for 1A4W precision, and 99.1%, 97.8% for 4A4W precision, for LeNet-5 architecture using MNIST and A-Z alphabet datasets respectively.

6G aims to achieve high-data rates while enhancing security. Terahertz (THz) technologies will be pivotal in 6G due to their higher frequencies and line-of-sight (LoS) propagation characteristics. However, these advantages can also pose challenges and expose THz communication to security risks. This paper thoroughly investigates the security challenges, explores potential solutions, and identifies existing research gaps in THz communications. The unique characteristics of the THz frequency band present significant hurdles, such as path loss, beam scattering, and vulnerability to interference and side-channel attacks. We discuss current mitigation strategies and emphasize the need for further research to bolster the security and reliability of terahertz communication systems.

The global distribution of the semiconductor supply chain has heightened the risk of hardware Trojans (HTs), small malicious circuits that adversaries may embed during various stages of the system-on-chip (SoC) design process. Frequently implanted by untrusted third parties, these HTs can operate covertly and, when activated, pose a serious threat to the integrity, performance, and functionality of the system. Although there are promising test generation techniques for HT detection, they face two significant practical limitations: a lack of scalability for large designs and insufficient trigger coverage. The effective detection of HTs requires the application of appropriate test vectors. This paper introduces PATROL, a novel algorithm that combines Particle Swarm Optimization (PSO) and Genetic Algorithms (GA) within a scalable framework for the detection of HTs. This framework employs Automated Test Pattern Generation (ATPG)based activation to achieve high trigger coverage. This approach significantly accelerates the convergence towards a solution while substantially improving the solution accuracy. Our experimental results demonstrate that our proposed method is more than 43 times faster and achieves an average increase in trigger coverage of more than 34%, significantly outperforming state-of-the-art test generation techniques for Trojan detection.

This paper introduces the MagNetX -an extension of the MagNet database to investigate transient hysteresis of magnetic materials. By employing automated data acquisition and measurement systems, detailed transient phenomena are observed and evaluated across a wide range of waveforms and operational conditions. Results reveal significant core loss variations during frequency transitions, and the B-H loop analysis further demonstrates loop area changes during transient phases. The platform enables advanced data-driven modeling of core losses and hysteresis loops over a wide range of amplitudes, frequencies, temperatures, and duty cycles under transient conditions, which are critical for practical circuit design when steady-state models are inadequate, such as modeling the magnetic core behaviors in ac-dc converters, motor drives, or power amplifiers.

This paper introduces a foundation neural network framework for modeling power magnetics in transient, based on MagNetX 1-a new extension of the MagNet database which includes extensive measurement data in transient. Provided with flux density B(t) and field intensity H(t) waveforms, the model uses partial memories and the next-state flux density excitation to predict the response of the field intensity in the next time step. The model is in time domain and is frequency independent. An example sequence-to-scalar LSTM neural network was designed, trained, and tested. This modeling framework can greatly enhance the modeling and design of power magnetics operating in transient condition, such as in PFCs and power amplifiers.

Today's trend in AR application customization focuses on addressing individual user needs, leveraging advancements in sensing technologies such as biosensors and eye trackers, alongside the evolution of generative AI models. These innovations facilitate the creation of dynamic and personalized content, including videos and images, tailored to specific models and rules. To address this evolving landscape, this research proposes a framework for Context-Aware AR Content Generation designed for cultural scenarios. This framework incorporates biosensors to monitor emotional states, eye-tracking to assess visual focus, and geospatial data to consider location and time. By integrating these inputs with generative AI, the framework not only generates and adapts culturally relevant AR content but also employs AR technology to provide historical context about the content elements on the user's device. Our framework aims to minimize the need for extensive user prompts and improve accessibility for individuals who may find textual or voice commands challenging.

Training data sets for unsupervised anomaly detection, expected to be anomaly-free, often contain anomalies (i.e., contamination) which can significantly degrade model performance. Most contamination-robust anomaly detection methods do not generalize across different architectures and rely on contamination ratio information which may not be available. In this paper, we propose a novel self-contained preprocessing anomaly detection approach focusing on data cleaning, i.e., identification and treatment of contamination within a training data set for a more effective unsupervised anomaly detection, all without relying on contamination ratio. Our approach is based on dynamic analysis of the cosine similarity between each data point and its reconstruction during training of an auto-encoder, anticipating low density in most anomaly regions within the dynamic space. This approach enhances data quality through hard cleaning: by removing all identified anomalous candidates from training data. Data cleaning allows generalization across various architectures as the improvements made at the data level benefit any anomaly detection model. This process yields up to 100% of contamination reduction, resulting average improvements of 11.2% in F1 score and 7.3% in accuracy across 8 models and 10 benchmark data sets.

Edge-AI applications face huge challenges in resource-constrained environments, particularly in enhancing computational efficiency within bandwidth limitations. This work proposes the Logarithmic-Posit-enabled Reconfigurable edge-AI Engine (LPRE) that enhances hardware efficiency without compromising accuracy. The proposed architecture utilizes timemultiplexed dynamically configurable single-layer hardware to balance resource reuse and bandwidth for multi-layer perceptron and CNN models. Evaluations on LeNet-5 using MNIST demonstrate that LPRE achieves up to 4× throughput enhancement at 8-bit precision with negligible accuracy loss (compared to FP32 baseline), while requiring up to 80% and 50% fewer resources than fixed-point arithmetic and state-of-the-art works, respectively. The design is viable for various edge-AI applications, such as real-time number plate recognition, offering scalable, energy-efficient IoT solutions.

Training data sets intended for unsupervised anomaly detection, typically presumed to be anomaly-free, often contain anomalies (or contamination), a challenge that significantly undermines model performance. Most robust unsupervised anomaly detection models rely on contamination ratio information to tackle contamination. However, in reality, the contamination ratio may be inaccurate. We investigate the impact of inaccurate contamination ratio information in robust unsupervised anomaly detection. We verify whether they are resilient to misinformed contamination ratios. It appears that there is a lack of discussion in the literature regarding the impact of incorrect contamination ratio information on unsupervised anomaly detection, despite its critical importance. Our investigation on 8 benchmark data sets reveals that such models are not adversely affected by exposure to misinformation. In fact, they can exhibit improved performance when provided with such inaccurate contamination ratios.

Deep auto-encoders (AEs) are widely employed deep learning methods in the field of anomaly detection, especially detection of security malicious activities. Cybersecurity analysts face challenges in managing a large volume of alerts, often constrained by limited processing resources. In response, various strategies, including false positive reduction, human-in-the-loop and alert prioritization, are employed. This paper explores the integration of uncertainty quantification (UQ) methods into alert prioritization for anomaly detection using an ensemble of AEs. UQ models recognize doubtful classification decisions, aiding analysts in treating the most certain alerts first (as a more certain alert is more likely to be accurate). Our study reveals a nuanced issue where applying UQ to an ensemble of AEs can lead to skewed distributions of large reconstruction errors, falsely indicating large standard deviation uncertainties on certain classification decisions. Contrary to intuition, AEs suggest that large reconstruction errors could indicate small uncertainties. To address this, we propose an extension for obtaining a calibrated standard deviation uncertainty distribution, mitigating erroneous alert prioritization. Evaluation on 6 benchmark intrusion detection datasets demonstrates that our proposed calibration approach enhances UQ methods' ability to prioritize alerts with a favorable trade-off in other invaluable performance metrics.

The Autonomous Robot Orchestration Solution (AROS) is transforming the management of robot fleets by identifying the state and environment of each robot and enabling them to collaborate toward common goals. This paper introduces AROS and its application in controlling massive fleets of Overhead Hoist Transport (OHT) vehicles in semiconductor fabrication facilities. AROS leverages key technologies, including reinforcement learning algorithms, discrete event simulation, and real-time data collection through Digital Twin (DT). The DT replicates the real system in a virtual environment with realtime communication to optimize decision-making for OHTs. A key innovation of AROS is the introduction of active Q routing, a dynamic routing method that adapts to changing traffic conditions by predicting and adjusting travel times through discrete event simulation. Active Q routing enhances operational efficiency by mitigating congestion and reducing delays, even in highly dynamic environments. We demonstrate the effectiveness of AROS and active Q routing on OHT system performance, showcasing reductions in average delivery times and increases in delivery capacity. These findings are validated through real-world use cases in a large-scale semiconductor fab.

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.

This work addresses the challenge of modelling Hammerstein nonlinear systems iteratively, combining a kernel adaptive filter with a linear filter. We evaluate this approach in the context of acoustic echo cancellation (AEC), where a nonlinear speaker followed by a linear system models the loudspeaker enclosure microphone system (LEMS). Our contributions include a novel iterative online processing procedure and a detailed investigation of challenges and strategies, such as offset removal and normalization, for improving system stability and performance. Our results demonstrate that normalization is crucial for reducing fluctuations and achieving a 5 dB improvement in Echo Return Loss Enhancement (ERLE) compared to linear filters.

This paper presents a cost-effective approach to enhancing multi-robot systems through the strategic integration of generative AI and machine learning capabilities. Building upon a traditional master-slave architecture utilizing Raspberry Pi and MSP430 microcontrollers, we introduce an optimized framework that significantly improves swarm robotics performance without requiring extensive hardware modifications. Our enhanced system implements distributed intelligence through federated learning, achieving a 17.9% improvement in object detection accuracy (from 76.4% to 94.3%) and a 65.3% reduction in system response latency (from 245ms to 85ms). The architecture incorporates adaptive model compression techniques and intelligent resource management, reducing communication overhead by 66.7% while maintaining model accuracy. Through comprehensive experimental validation, we demonstrate that the enhanced system achieves a 31.2% reduction in task completion time while operating within existing hardware constraints. The implementation maintains cost-effectiveness with an enhancement cost of $86.37 per unit and achieves return on investment within 1.9 months under standard operational conditions. Our results establish that sophisticated swarm behaviors can be achieved through software optimization rather than hardware upgrades, providing a practical pathway for upgrading existing robotic infrastructure in industrial applications.

The integration of artificial intelligence (AI) within Open Radio Access Network (Open RAN) architectures represents a fundamental transformation in telecommunications infrastructure. This paper presents a comprehensive analysis of AI integration in Open RAN, focusing on its implications for 6G networks and beyond. We examine the effectiveness of various AI models in optimizing network performance, comparing traditional optimization techniques with AI-driven approaches through extensive simulations and real-world deployments. Our results demonstrate significant improvements, including a 23% enhancement in resource utilization efficiency and a 27% reduction in power consumption. Through case studies of commercial deployments, we validate the feasibility and benefits of AI-enhanced Open RAN systems. The research also addresses critical challenges in standardization, security, and multi-vendor integration, proposing practical solutions for implementation. Finally, we provide recommendations for practical deployment and identify promising directions for future research in this rapidly evolving field.

 Reconfigurable Intelligent Surfaces (RIS) represent a transformative technology for achieving the ambitious performance requirements of 6G wireless networks. This comprehensive review examines advanced beamforming techniques incorporating RIS technology, presenting a systematic analysis of theoretical foundations, implementation challenges, and future directions. We develop a rigorous analytical framework that characterizes the fundamental performance limits of RIS-assisted beamforming, demonstrating that system capacity scales as O(log N) with the number of reflecting elements while maintaining energy efficiency that follows O(log N/N). Our analysis reveals that hybrid approaches combining traditional optimization with machine learning techniques can achieve 85-90% of theoretical performance bounds while reducing computational complexity by approximately 80%. We present a detailed examination of practical implementation challenges, including hardware impairments, channel estimation overhead, and real-time optimization requirements. The study establishes that RIS-assisted beamforming can enable a 300% increase in spectral efficiency and 60-70% reduction in power consumption compared to conventional massive MIMO systems. Additionally, we identify critical research directions, including distributed RIS optimization, multi-user scheduling, and integration with artificial intelligence frameworks. This review provides a comprehensive foundation for researchers and practitioners working toward the realization of RIS-enabled 6G networks, offering both theoretical insights and practical implementation guidelines.

Teacher education is essential in every stage of a teacher's career, especially at the start as it provides a strong foundation in pedagogy, theoretical knowledge, and reflective attitudes. In this paper, a custom OpenAI GPT is presented that generates educational scenarios with possible students' questions and gives feedback to science teachers' answers and educational approaches. Furthermore, a pilot study was carried out that measured the self-concept of the science teachers and showed that the influence of the aforementioned GPT-teachers interaction had a positive impact on teachers' self-concept, indicating an interesting way generative AI tools can be a valuable educational partner.