To better comprehend the upper-limb needs and challenges stroke survivors face, and the issues associated with existing technology, the authors conversed with 153 members of the ecosystem (60 neuro patients, 30 caregivers, and 63 medical providers). Patients fell into two populations depending on their upper-limb impairment: spastic (stiff, clenched hands) and flaccid (limp hands). The authors chose to focus on the second category and developed a set of design constraints based on the information collected through customer discovery. With these in mind, they designed and prototyped OrthoHand Flex, a 3D-printed powered wrist-hand grasping orthosis (exoskeleton) to aid in recovery. The orthosis is easily custom-sized based on two parameters and derived anatomical relationships. The researchers tested the prototype on a survivor of stroke and modeled the kinematic behavior of the orthosis with and without load. The prototype neared or exceeded the target design constraints and was able to grasp objects consistently and stably, as well as exercise the patientsâ\euro™ hands. In particular, donning time was only 42 seconds, as compared to the next fastest time of 7 minutes reported in literature. This device has the potential to revolutionize neurorehabilitation by making assistive devices accessible to patients at home, and it lays the foundation for clinical trials and further device development

In this paper, we present a novel digital-like defect-oriented Built-In Self-Test (BIST) methodology for Analog and Mixed-Signal (AMS) circuits. The core idea of this approach centers around the segmentation of complex AMS circuits into smaller, more manageable units for analysis. Emphasizing resource utilization efficiency, we highlight the necessity of employing purely digital circuits for both injectors and monitors within the BIST framework. We demonstrate the effectiveness of this approach through the development of a BIST system for a 12-bit Successive Approximation Register Analog-to-Digital Converter (SAR ADC). Notably, our methodology achieves 100% defect coverage without introducing additional BIST circuitry for subcircuit testing, relying solely on digital monitors for sampling switch evaluation. Furthermore, our proposed approach incurs minimal area overhead, resulting in a fast and comprehensive defect-oriented BIST solution. This versatile test method can be deployed post-manufacturing or in-field, offering flexibility in its application timing.

This research paper titled “Enhanced Wildfire Detection using AI/ML: Harnessing Multi-spectral Satellite Imagery with Convolutional Neural Networks” aims to advance the capabilities of wildfire detection by employing Artificial Intelligence (AI), specifically Convolutional Neural Networks (CNNs). Given the escalating threat of wildfires exacerbated by climate change and human activity, traditional detection methods, though effective, are both costly and time-consuming. To counter these limitations, the study taps into multi-resolution satellite imagery, particularly from the VIIRS and Sentinel-2 satellites. The primary data source, VIIRS, offers comprehensive spectral bands and frequent global coverage. In contrast, Sentinel-2 provides high-resolution optical image data vital for detailed wildfire detection. The research processes the collected data, refining and categorizing them for training and testing. A Convolutional Neural Network is then employed to classify images as either “fire” or “nofire.” Two main architectures, Deep CNN and a simplified MobileNet-like CNN, were explored. Among the models tested, the Deep CNN using the Adam optimizer was found to be the most accurate, although it hinted at possible overfitting. The paper also points out several limitations, such as reliance on the visible spectrum that could be obstructed by atmospheric conditions and the temporal gaps in image captures that could delay real-time detection. The study concludes by emphasizing the transformative potential of integrating AI with satellite technology for early wildfire detection. Future advancements could harness multispectral bands and refine spatial and temporal resolutions to further enhance the early detection and intervention of wildfires. The research received support from the Network of Resources (NoR) at ESA, which facilitated expanded access to the SentinelHub platform.

Anomaly detection (AD) in medical images aims to recognize test-time abnormal inputs according to normal samples in the training set.  Knowledge distillation based on the teacher-student (T-S) model is a simple and effective method to identify anomalies, yet its efficacy  is constrained by the similarity between teacher and student network architectures. To address this problem, in this paper, we propose a T-S model with skip connections (Skip-TS) which is trained by direct reverse knowledge distillation (DRKD) for AD in medical images. First, to overcome the low sensitivity to anomalies caused by structural similarity, we design an encoder-decoder architecture where the teacher network (T-Net) is a pre-trained encoder and the student network (S-Net) is a randomly initialized decoder. During training, the S-Net learns to reconstruct the shallow representations of images from the output of the T-Net, which is called DRKD. Secondly, we introduce skip connections to the T-S model to prevent the S-Net from missing normal information of images at multi-scale. In addition, we design a multi-scale anomaly consistency (MAC) loss to improve the anomaly detection and localization performance. Thorough experiments conducted on twelve public medical datasets and two private medical datasets demonstrate that our approach surpasses the current state-of-the-art by 6.4% and 8.2% in terms of AUROC on public and private datasets, respectively. Code and organized benchmark datasets will be available at https://github.com/Arktis2022/Skip-TS.

In the ever-evolving field of technologies, the emergence of Artificial General Intelligence (AGI), often referred as strong artificial intelligence (AI), stands as a breakthrough in the realm of machines intelligence, promising to witness a new era of capabilities and possibilities. In particular, AGI ventures into human-level cognition, and expands to thinking, reasoning, and awareness. This imminent evolution is envisioned to be manifested through the embodiment of AI machines, allowing machines to transcend their purely computational nature and interact with the world through the different senses. Accordingly, AI agents will be grounded in the physical environment, going through subjective experiences and acquiring the needed knowledge that will lead to understanding and cognition. In our article, we explore the path towards realizing the true vision of AGI through AI embodiment, where we dig into the different types of thinking required to achieve knowledge, and hence, cognition and understanding. Furthermore, we look through the evolution of generative AI models, and shed lights on the limitations of auto-regression in large language models (LLMs), with the aim to answer the question: is sensory grounding (through 6G) necessary, and enough, to achieve understanding in LLMs? Finally, we identify the main pillars of AGI and unveil how 6G networks will orchestrate the development of AGI systems.

A brief but holistic overview about THz communications and Sensing in the upcoming 6g and beyond wireless communications systems. It highlights the major opportunities and challenges to be addressed.

Handwritten paper charting remains the primary method of collecting anesthesia health data in low-middle income countries that lack automated data capture devices and electronic medical records. Lack of digital data precludes automated evaluation of a patientâ\euro™s condition in real time intraoperatively and limits healthcare research to improve outcomes. In this paper, we demonstrate the use of computer vision software to digitize handwritten intraoperative data from smartphone photographs of paper anesthesia records from the University Teaching Hospital of Kigali. We implement approaches for removing perspective distortions from photographs, removing shadows and improving image readability though morphological operations, reading handwritten symbols using deep neural networks, and decoding handwritten symbols into meaningful values. Our work builds upon the contributions of previous research by improving upon their methods, updating the deep learning models to newer architectures, as well as consolidating them into a single piece of software. Our software digitizes checkbox data with greater than 99% accuracy and digitizes blood pressure data with a mean average error of 1.17mmHg under realistic photography conditions, and digitizes physiological data to within human accuracy when provided legible handwriting and high image quality. Our contributions provides improved access to digital data to healthcare practitioners in low-middle income countries and enables collection of data previously inaccessible to clinicians and researchers.Â

Human-vehicle co-pilot has recently been the research focus in autonomous vehicles. However, there have been many subjective uncertainties in the driver’s state and the behavior of human-vehicle interaction, which is a crucial factor constraining the safety of Human-in-the-loop co-driving. To promote the research in the domain, we propose a new visual?tactile perception method based on a driving simulation platform and have built a new large-scale dataset. The data of drivers’ visual and tactile behaviors and states during driver fatigue and human-vehicle switched driving are included to satisfy different investigation needs. Concerning visual data, we extract features from the drivers’ behaviors and states through a deep convolutional network in multiple angles, followed by adjusting the model with subjective feedback from Human-in-the-loop to improve the accuracy of the extracted feature through continuous iterations. Regarding haptic processing, we designed a steering wheel using flexible conductive electronic-skin material, which achieved time-series data extraction built upon the driver’s heart rate and rate of change based on the haptic signal when the drivers are holding the steering wheel. Moreover, we calibrated the extracted data with the wearable heart rate sensor. Finally, we synchronized the time-series visual-tactile data with time stamps, forming a cross-modal database that provides data support for implementing cross-modal perception algorithms for driver behavior. VTD Dataset Download Address： VTD Datasets (merakoswang.github.io)Â

Ensuring the security and reliability of cooperative vehicle-to-vehicle (V2V) communications is an extremely challenging task, due to the dynamic nature of vehicular networks as well as the delay-sensitive wireless medium. The moving target defense (MTD) paradigm has been proposed to overcome the challenges of conventional solutions, based on static network services and configurations. Specifically, the MTD approach involves the dynamic altering of the network configurations to improve resilience to cyberattacks. Nevertheless, the current MTD solution for cooperative networks has several limitations, such as they are not well-suited for highly dynamic environments; they require high synchronization modules that are resource-intensive and difficult to implement; and finally, they rely heavily on the attack-defense models, which may not always be accurate or comprehensive to use. In this paper, we propose an intelligent spatiotemporal diversification MTD scheme to defend against eavesdropping attacks in cooperative V2V networks. Specifically, we design benign random data injection patterns to meet the security and reliability requirements of the vehicular network. Our methodology involves modeling the configuration of vehicular relays and data injection patterns as a Markov decision process, followed by applying deep reinforcement learning to determine the optimal configuration. We then iteratively evaluate the intercept probability and the percentage of transmitted real data for each configuration until convergence is achieved. In order to optimize the security-real data percentage (S-RDP), we developed a two-agent framework, namely MTD-DQN-RSS & MTD-DQN-RSS-RDP. The first agent, MTD-DQN-RSS, tries to minimize the intercept probability by injecting additional fake data, which in turn reduces the overall RDP, while the second agent, MTD-DQN-RSS-RDP, attempts to inject a sufficient amount of fake data to achieve a target S-RDP. Finally, extensive simulation results are conducted to demonstrate the effectiveness of our proposed solution where they improved the system security by almost 28% and 49%, respectively compared to the conventional relay selection approach.

The excellent generalization, contextual learning, and emergence abilities in the pre-trained large models (PLMs) handle specific tasks without direct training data, making them the better foundation models in the adversarial domain adaptation (ADA) methods to transfer knowledge learned from the source domain to target domains. However, existing ADA methods fail to account for the confounder properly, which is the root cause of the source data distribution that differs from the target domains. This study proposes an adversarial domain adaptation with confounder balancing for PLMs fine-tuning (ADA-CBF). The ADA-CBF includes a PLM as the foundation model for a feature extractor, a domain classifier and a confounder classifier, and they are jointly trained with an adversarial loss. This loss is designed to improve the domain-invariant representation learning by diluting the discrimination in the domain classifier. At the same time, the adversarial loss also balances the confounder distribution among source and unmeasured domains in training. Compared to existing ADA methods, ADA-CBF can correctly identify confounders in domain-invariant features, thereby eliminating the confounder biases in the extracted features from PLMs. The confounder classifier in ADA-CBF is designed as a plug-and-play and can be applied in the confounder measurable, unmeasurable, or partially measurable environments. Empirical results on natural language processing and computer vision downstream tasks show that ADA-CBF outperforms the newest GPT-4, LLaMA2, ViT and ADA methods.

 Microwave photonic (MWP) signal processors, which process microwave signals based on photonic technologies, bring advantages intrinsic to photonics such as low loss, large processing bandwidth, and strong immunity to electromagnetic interference. Optical microcombs can offer a large number of wavelength channels and compact device footprints, which make them powerful multi-wavelength sources for MWP signal processors to realize a variety of processing functions. In this paper, we experimentally demonstrate the capability of microcomb-based MWP signal processors to handle diverse input signal waveforms. In addition, we quantify the processing accuracy for different input signal waveforms, including Gaussian, triangle, parabolic, super Gaussian, and nearly square waveforms. Finally, we analyze the factors contributing to the difference in the processing accuracy among the different input waveforms, and our theoretical analysis well elucidates the experimental results. These results provide a guidance for microcomb-based MWP signal processors when processing microwave signals of various waveforms.Â

Attacks and anomalies are quiet common in todayâ\euro™s technological world. However it is important to identify these anomalies as soon as possible since these anomalies can cause various problem. Therefore anomaly detection plays a vital role in detecting these outliers. Anomaly detection detects suspicious activity which fundamentally falls outside of the respected pattern of behaviour. Anomaly detection aims at identifying the unpredicted and unexpected events in data streams which are commonly referred to as anomalous events. Therefore proposed modified linearity based grey swarm algorithm and hyperparameter optimization techniques for improving the performance of the proposed model. Hyperparameter optimization techniques implemented in the proposed model are genetic algorithm (GA), Grid Search and Random Search. Feature selection is implemented in the proposed model is modified linearity based grey swarm algorithm since it has the ability to reduce the convergence. In addition to modified linearity based grey swarm algorithm, modified objective function is employed in the proposed model as it has the capability to find solutions to problems which are difficult to solve. Finally the performance of the proposed model is evaluated and assessed using various performance metrics such as accuracy, recall, precision and F1 score. At last, proposed model is compared with the existing models and found that proposed model outperformed other existing models.

This paper introduces Continual Learning for Multilingual ASR (CL-MASR), a benchmark for continual learning applied to multilingual ASR. CL-MASR offers a curated selection of medium/low-resource languages, a modular and flexible platform for executing and evaluating various CL methods on top of existing large-scale pretrained multilingual ASR models such as Whisper and AWavLM, and a standardized set of evaluation metrics.

We study the variations in the coherence between breathing & EEG as a function of breathing rate, without and with breath hold. Â

This review focuses on the integration of intelligent driving and intelligent cockpit systems. With the advancement of intelligent levels, autonomous vehicles are capable of driving in complex driving context. However, the lack of systems that consider both driver and driving scenarios information reduces the success probability of decision-making of autonomous vehicles in driving context where driver and driving scenarios information are tightly coupled during the human-vehicle cooperation stage. To solve this problem, we present the Cockpit-Driving Integration system (CDI) and review perception and decision-making algorithms for CDI systems. Additionally, to achieve human-centric autonomous vehicles, we propose that the CDI system should consider the personalized characteristics of drivers. Finally, we present a framework for CDI systemsÂ

This piece of research identifies how DTW and some common environmental parameters influence crop yield or vegetation. A total of 14 DTW and its variants and 10 environmental parameters that promote crop yield or vegetation were adopted. The work points out the potential gaps untapped within DTW application in crop yield prediction and the efficacy of this model and its variants in forecasting yield.

In this letter, we consider a secondary network (SN) in which an ambient backscatter device (BD) utilizes the secondary transmitter (ST) signal to communicate its own information to the secondary destination (SD). Optimizing performance of such networks is complicated by signal reflections by the BD. It is shown in this work how the secondary transmit power and the reflection coefficient of the BD can both be jointly optimized using a simple restricted one dimensional search to satisfy the quality of service (QoS) constraints of SN as well as the primary network (PN) while maximizing performance of the backscatter link, which is termed the tertiary network (TN). Only statistical channel knowledge is used for this purpose. It is seen that careful optimization can improve spectral efficiency. Simulations validate the derived analytical expressions.

A document by Aristeidis Seretis . Click on the document to view its contents.