Search

The use of 3D point clouds (3DPCs) in deep learning (DL) has recently gained popularity due to several applications in fields such as computer vision, autonomous systems, and robotics. DL, as a dominant artificial intelligence approach, has been effectively applied to handle a variety of 3D vision challenges. However, building strong discriminative feature representations from irregular and unordered PCs is difficult. Self-driving systems commonly employ lidar to obtain precise 3D geometric data around vehicles for perception, path planning, and localization. The semantic segmentation of lidar-based PCs is a key activity that must be completed in real time. However, the majority of current convolutional neural network models for 3DPC semantic segmentation are extremely complex and cannot be processed in real time on an embedded platform. In this article, our goal is to offer a comprehensive review of current advances in DL approaches for PC feature representation, including 3DPC segmentation, and future challenges. [ABSTRACT FROM AUTHOR]

Real-time monitoring of 1-D biopotentials, such as electrocardiograms (ECG), necessitates effective feature extraction and classification, a strength of deep learning (DL) algorithms. Designing 1-D convolutional neural network (1-D CNN) accelerators for biopotential classification via open-source codesign workflows, particularly high-level synthesis for machine learning (HLS4ML), offers advantages over GPU-based or cloud-based solutions, including high performance, low latency, low power consumption, swift time-to-market, and cost-effectiveness. We present an implementation of a quantized-pruned (QP) 1-D CNN model on the PYNQ Z2 SoC using HLS4ML by seamlessly deploying its soft IP core generated via Vivado Accelerator backend, showcasing the efficacy of quantization-aware training (QAT) in reducing power consumption to 1.655 W from 1.823 W. Our approach demonstrates improved area consumption, resource utilization, and inferences per second compared to the baseline (B) 1-D CNN model, with a controlled 4% or less reduction in weighted Accuracy, Precision, Recall, and F1-score, revealing the nuanced tradeoffs between performance metrics and system efficiency for real-time 3-channel ECG Arrhythmia classification. [ABSTRACT FROM AUTHOR]

Fog devices in fog computing frameworks are responsible for fetching and executing the tasks submitted by the deployed resource-constraint embedded edge devices. Based on the availability of resources, tasks are offloaded to the virtual machines hosted by the fog devices. These tasks may then get scheduled to guarantee a number of efficiency-related metrics. While throughput has a decisive impact on the timely execution of tasks, the appropriate utilization of this metric has not been considered in the existing mechanisms. In this letter, we first discuss the proper use of this objective in the fitness function of meta-heuristic algorithms. Then, we explain that adopting throughput by the fitness functions in the form of two conventionally used weighted-sum, and fractional techniques may ignore solutions with a better guarantee ratio. Consequently, we propose a novel approach called DATA to be replaced with these two old approaches. DATA is a throughput, and deadline-aware task scheduling mechanism for time-sensitive fog frameworks, which its fitness function utilizes genetic optimization by encoding the solutions into chromosomes. It uses single-gene mutation and two-point crossover. In this approach, two populations are considered to search the problem space. The main population is evaluated based on the guarantee ratio, while the helper population is evaluated based on the throughput. Furthermore, the helper population uses weighted-sum. The initial population is generated randomly by the uniform distribution, to provide a load-balancing. Based on our extensive evaluations, the selected solution by DATA provides the highest guarantee ratio, while having the lowest possible makespan. [ABSTRACT FROM AUTHOR]

Digital twins of the semiconductor fabrication process provide means for optimization of the physical layout and nanofabrication process design, studying compatibility between desired structures and a process flow, and a pathway to analyze the root causes of defects for state-of-the-art CMOS and MEMS devices. In this paper, a metric for the geometric differences between structures visualized by CD-SEM images is defined, and a computer-vision-based algorithm is developed to evaluate the metric. One of the major uses of such metrics is to compare experimental and simulated images. For this application, numerical results are presented when the simulator is SEMulator3D®, a physics-based process modeling software system for semiconductor and MEMS devices. Computer vision tools, such as filters, thresholding, and morphology operations, are used to extract geometric features from CD-SEM images and pattern matching and symmetric difference are used to compute the metric. Examples of using the metrics to quantify the geometric similarity between a simulated nanostructure and an experimental CD-SEM image of the fabricated nanostructure are presented. The data consists of eight classes of nanostructures which are defined, fabricated in the cleanroom with 36 combinations of layout parameters, and imaged with a CD-SEM. [ABSTRACT FROM AUTHOR]

Despite the recent advancements in dialogue systems, persona-driven chatbots are still in their infancy. Previous studies on persona-driven dialogue generation demonstrated its ability in generating responses that contain more detailed persona information. However, the challenge of maintaining persona consistency and contextual coherence still persists in persona-driven dialogue generation. Moreover, current methods have limitations in processing multi-source inputs and identifying interlocutor intents due to the absence of trustworthy labels and effective modeling. Additionally, numerous approaches rely on pre-trained large-scale language models that require costly computational resources. To address these challenges, a lightweight hierarchical intent-inferring pointer network is proposed for multi-source persona-driven dialogue generation. The proposed method involves detecting interlocutor intents in chitchat and utilizing pseudo labeling and natural language inference techniques to generate intent labels. Our model is evaluated on a benchmark dataset PersonaChat. The experimental results show that our model outperforms the strongest baseline by 13.47% and 4.28% in terms of persona consistency and contextual coherence, respectively. [ABSTRACT FROM AUTHOR]

The t-distributed stochastic neighbor embedding (t-SNE) consists of a powerful algorithm for visualizing high-dimensional data in a lower dimensional space. It is extensively employed in machine learning (ML) and data analysis, including unsupervised metric learning. In this article, we propose improvements concerning two main aspects of the t-SNE. First, the incorporation of class labels is adopted to increase its suitability for supervised classification. Second, stochastic and geodesic distances are used as dissimilarity measures to avoid the dependence of the standard Euclidean distance, which is particularly sensitive to outliers. Computational experiments with several real-world datasets indicate that the proposed methodological approach is capable of improving classification accuracy compared with established methods. The results indicate a superior performance compared with the regular t-SNE and linear discriminant analysis (LDA), and a dependence on fewer parameters in comparison with the state-of-the-art supervised uniform manifold approximation and projection (UMAP) algorithm. [ABSTRACT FROM AUTHOR]

To address the challenges of efficient hardware design for error-tolerant applications, several techniques of applied approximate computing have been proposed. Pruning algorithms aim to approximate circuits with reduced design requirements at the cost of an acceptable degradation of their quality of result. In this letter, we present the effects of resynthesis, an iterative application of logic synthesis along with pruning algorithms, into a state-of-the-art approximate design flow, AxLS. Resynthesis strategy improves the approximation, achieving up to 70% area-power savings for the same error in the output, and reducing the number of iterations, and hence the time required to explore the design space in up to $30\times $ , to obtain an approximated design. [ABSTRACT FROM AUTHOR]

Heart sound auscultation has been applied in clinical usage for early screening of cardiovascular diseases. Due to the high demand for auscultation expertise, automatic auscultation can help with auxiliary diagnosis and reduce the burden of training professional clinicians. Nevertheless, there is a limit to classic machine learning's performance improvement in the era of Big Data. Deep learning has outperformed classic machine learning in many research fields, as it employs more complex model architectures with a stronger capability of extracting effective representations. Moreover, it has been successfully applied to heart sound analysis in the past years. As most review works about heart sound analysis were carried out before 2017, the present survey is the first to work on a comprehensive overview to summarise papers on heart sound analysis with deep learning published in 2017–2022. This work introduces both classic machine learning and deep learning for comparison, and further offer insights about the advances and future research directions in deep learning for heart sound analysis. [ABSTRACT FROM AUTHOR]

Computer mouse tracking is a simple and cost-efficient way to gather continuous behavioral data. As theory suggests a relationship between affect and sensorimotor processes, the computer mouse might be usable for affect sensing. However, the processes underlying a connection between mouse usage and affect are complex, hitherto empirical evidence is ambiguous, and the research area lacks longitudinal studies. The present work brings forward a longitudinal field study in which 179 participants hourly self-reported their affect while their mouse usage was tracked both during their self-directed, contextless as well as task-bound computer use over the course of 14 days, resulting in a dataset comprising 10,760 instances of data collection. Extensive statistical analysis using null hypothesis significance testing and machine learning reveal weak and sporadic relationships between mouse usage and longitudinal self-reported affect at best. The results of this study challenge the use of computer mouse tracking for longitudinal affect sensing and point to a necessity for more research. [ABSTRACT FROM AUTHOR]

EEG emotion recognition has been hampered by the clear individual differences in the electroencephalogram (EEG). Nowadays, domain adaptation is a good way to deal with this issue because it aligns the distribution of data across subjects. However, the performance for EEG emotion recognition is limited by the existing research, which mainly focuses on the global alignment between the source domain and the target domain and ignores much fine-grained information. In this study, we propose a method called Graph-based Unsupervised Subdomain Adaptation (Gusa), which simultaneously aligns the distribution between the source and target domains in a fine-grained way from both the channel and emotion subdomains. Gusa employs three modules, such as the Node-wise Domain Constraints Module to align each EEG channel and obtain a domain-variant representation, the Class-level Distribution Constraints Module, and the Emotion-wise Domain Constraints Module, to collect more fine-grained information, create more discriminative representations for each emotion, and lessen the impact of noisy emotion labels. The studies on the SEED, SEED-IV, and MPED datasets demonstrate that Gusa significantly improves the ability of EEG to recognize emotions and can extract more granular and discriminative representations for EEG. [ABSTRACT FROM AUTHOR]

High-level synthesis (HLS) has democratized field programmable gate arrays (FPGAs) by enabling high-level device programmability and rapid microarchitecture customization through the use of directives. Nevertheless, the manual selection of the appropriate directives, i.e., the annotations included in the high-level source code to instruct the synthesis process, is a difficult task for programmers without a hardware background. In this letter, we present CollectiveHLS, an ultrafast knowledge-based HLS design optimization method that automatically extracts the most promising directive configurations and applies them to the original source code. The proposed optimization scheme is a fully data-driven approach for generalized HLS tuning, as it is not based on quality of result models or meta-heuristics. We design, implement, and evaluate our method with more than 100 applications of Machsuite, Rodinia, and GitHub on a ZCU104 FPGA. We achieve an average geometric mean speedup of x14.1 and x10.5 compared to the unoptimized, i.e., without HLS directives and optimized designs, a high design feasibility score, and an average inference latency of 38 ms. [ABSTRACT FROM AUTHOR]

Machine learning models play a crucial role in sports monitoring by effectively identifying various activities and tracking the number of repetitions during repetitive movements. However, creating models that accurately detect different types of exercises and provide feedback on movement adjustments for wearable devices remains a challenge. In this letter, we propose an alternative approach that addresses this issue by using the restricted Boltzmann machine (RBM) algorithm to learn, evaluate, and provide adjustment feedback based on inertial sensor data in real-time. Our experimental results show that by evaluating body segments individually, highly specialized models can be generated from a small set of movement repetitions. Moreover, these models have the capability to offer users precise recommendations on how to fine-tune the intensity, acceleration, and amplitude of the monitored segment. By using our proposed method, there is a great potential to enhance the accuracy and effectiveness of wearable devices used for sports monitoring. [ABSTRACT FROM AUTHOR]

The air ventilation of enclosed premises has a direct impact on the occupants’ well-being. If not properly regulated, the air ventilation can originate a multitude of diseases and pathologies. The present study proposes a new KPI (ventilation KPI) adapted to Smart Cities. It is especially designed for academic environments (Smart Universities) in which community members spend a long time gathered in classrooms, seminars, laboratories, etc. The ventilation KPI (or KPIv) was designed to support decision making and is based on the estimation of the number of occupants of an enclosed space and the accumulation of existing CO2. Two AI techniques are proposed to perform these estimations, specifically, two regressive neural networks. The resulting models, together with the KPI, were implemented through the development of value-added services for the University of Alicante’s Smart University platform. The network models were designed to be embedded within the built IoT device prototypes. These prototypes are small and inexpensive. They act as intelligent sensors and are connected via a low consumption and emission network (LoRa). The case study showed that it is possible to take advantage of the preexisting services and resources of these platforms and to validate the KPIv. [ABSTRACT FROM AUTHOR]

The harmonic grid impedance is a crucial system property utilized in various applications. However, traditional simulation-based methods are often limited by the lack of access to essential information, such as equipment characteristics, load behaviors, and grid topologies. In response, measurement-based black-box techniques have been developed, with initial commercial solutions available for low-voltage applications. Unfortunately, for the medium-voltage level, only a limited number of prototypes exist. Unlike other techniques, which require large, immobile setups, the measurement system described in measurement system is the first of its kind to be successfully applied in several research projects at various grid connection points. This paper presents, for the first time, real-world grid impedance measurements collected from multiple sites over extended time periods on the medium-voltage level. In addition to the measurement data, this paper provides comprehensive metadata for each measurement site, which is presented in measurement sites and metadata and the appendix, significantly enhancing the usability of these results for further research. The measurement results have been thoroughly analyzed, providing valuable insights into medium-voltage grid behavior. The evaluation of the measurement data in presentation and evaluation of the measurement data includes general observations comparing grid impedances at various connection points, as well as graphical and statistical evaluations of the time-dependent behavior of grid impedance. Special attention is given to resonance points, and the influence of renewable energy installations on grid impedance. Finally, the results are summarized and discussed in conclusion of the measurement results.

In this study, we conducted a comprehensive longitudinal measurement study of domain names associated with major global events. We aimed to understand the registrants’ motives, usage, and abuse of these domain names. We specifically focused on the Olympic Games since they attract sustained attention from when the venue is announced to the event’s conclusion. Our study focused on the Tokyo, Beijing, and Paris Olympics. Our three-year investigation revealed that the number of Olympic-related domain name (ODN) registrations increased concurrently with the postponement of the 2020 Tokyo Olympics and the diplomatic boycott of the 2022 Beijing Olympics. Furthermore, we discovered a substantial increase in the number of ODNs used for malicious websites just before the games. Many ODNs related to the regional nature of each game were acquired, and several ODNs required close attention from a security perspective.

Time-Sensitive Networking (TSN) encompasses standards to facilitate near-deterministic performance in wired networks and the integration of TSN with modern wireless systems, such as Fifth Generation (5G) systems, is intensely pursued. Consequently, TSN is examined by numerous research groups. Given the limited and costly access to TSN hardware, a reliable simulator becomes crucial. OMNeT++ is a modular simulator that can be expanded with network simulation models, such as INET, which can simulate TSN functionalities. We evaluate the accuracy of the OMNeT++ INET simulator in mirroring the physical (hardware-based) reality by comparing OMNeT++ INET simulations with measurements of two commercial TSN switches. Our evaluations encompass the generalized Precision Time Protocol (gPTP) accuracy, the Store-and-Forward (SF) and Cut-Through (CT) switching forwarding latencies, the Time Aware Shaper (TAS), and Frame Preemption (FP), including combinations of CT, TAS, and FP. We find that compared to hardware measurements, INET simulations exhibit different clock synchronization dynamics, underestimate the switching latencies, and do not support CT with FP. We enhance the alignment of INET simulations with authentic hardware behaviors by modifying INET modules. We modify the INET gPTP synchronization model and the INET simulation models for the SF and CT forwarding latencies. Also, we modify the INET simulation modules to support the combined operation of CT and FP. We demonstrate how our INET modifications, which we make publicly available, accurately simulate the behaviors of real TSN hardware.

Ensuring quality human-AI interaction (HAII) in safety-critical industries is essential. Failure to do so can lead to catastrophic and deadly consequences. Despite this urgency, existing research on HAII is limited, fragmented, and inconsistent. We present here a survey of that literature and recommendations for research best practices that should improve the field. We divided our investigation into the following areas: 1) terms used to describe HAII, 2) primary roles of AI-enabled systems, 3) factors that influence HAII, and 4) how HAII is measured. Additionally, we described the capabilities and maturity of the AI-enabled systems used in safety-critical industries discussed in these articles. We found that no single term is used across the literature to describe HAII and some terms have multiple meanings. According to our literature, seven factors influence HAII: user characteristics (e.g., user personality), user perceptions and attitudes (e.g., user biases), user expectations and experience (e.g., mismatched user expectations and experience), AI interface and features (e.g., interactive design), AI output (e.g., perceived accuracy), explainability and interpretability (e.g., level of detail, user understanding), and usage of AI (e.g., heterogeneity of environments). HAII is most measured with user-related subjective metrics (e.g., user perceptions, trust, and attitudes), and AI-assisted decision-making is the most common primary role of AI-enabled systems. Based on this review, we conclude that there are substantial research gaps in HAII. Researchers and developers need to codify HAII terminology, involve users throughout the AI lifecycle (especially during development), and tailor HAII in safety-critical industries to the users and environments.

Globally, new technological solutions are being studied and developed by researchers in the field of power and energy systems. Before these solutions can be integrated into society as marketable products, thorough testing and validation of the design must be conducted. The National Center for Reliable Electric Power Transmission (NCREPT) is a unique high-power test facility located at the University of Arkansas, Fayetteville, AR, USA. NCREPT is a 12,000-square-foot laboratory space capable of testing power levels up to 6 MVA and offers testing at both the 480 V and 13.8 kV levels. NCREPT's capabilities expand outside of the scope of simply high-power testing. NCREPT has also supported experimental testing for transportation electrification, hardware-in-the-loop, and cybersecurity applications. NCREPT has personnel with experience testing new technologies and has demonstrated its innovative approaches by being a part of four R&D 100 award-winning projects. This paper gives an extensive overview of NCREPT, highlighting its significance as a state-of-the-art testing laboratory.

Food texture is an essential factor in the perception of chewing. Compared to taste and aroma, food texture is dominant in the palatability of solid and semi-solid foods. Hence, food development processes require a method for evaluating broad food textures. This study proposes a prediction method for food texture using multiple measurements and template data. First, a measurement system recorded the force, vibration, and sound pressure data during food compression. The moisture rate of food was also measured by a moisture meter. Second, many template data are automatically determined from the outline waveforms of measurement data. Third, the dynamic time warping calculates distance vectors between measurement and template data. Finally, the Gaussian process regression algorithm determines the relationship between the distance vectors and sensory evaluation data. The advantage of using template data is that there is no need to extract specific features from measurement data. The effectiveness of the proposed method was validated through sensory evaluation and measurement experiments. The proposed method was able to predict food texture value with low errors through the experiment with nine textures and 21 samples.

Blockchain’s promise of an immutable distributed ledger, fully fault tolerant with no centralized control deployed in an untrusted decentralized peer-to-peer network has disrupted the financial and technical industry. Blockchains may vary in terms of their design goals and algorithms, but they all rely on one common premise: the blockchain network is a decentralized peer-to-peer system and the ledger is stored on these untrusted peers. In this paper, we measure the characteristics of blockchain peers in the top four blockchains with dominant market share, Bitcoin, Ethereum, Ripple, and Solana. All four blockchains have a common trend that they have moved away from the idealistic decentralized peer-to-peer systems in their original design, and have become more server-like, more centralized, and more vulnerable to certain attacks. Using our measurements, we quantify the amount of effort required to delay, disrupt, and attack these blockchains.

This paper proposes a process for measuring the rated power electrical characteristics of inductive power transfer (IPT) coupling pads in limited laboratory environments through topology reconfiguration. Among the components of IPT systems, the coupling pad is responsible for the main losses in the converter. Moreover, coupling pads have nonlinear characteristics that depend on various factors, such as the number of coil turns, the diameter, the permeability of the magnetic material, and the amount of aluminum. Therefore, verifying the operation is necessary when applying various position and control algorithms after configuring an IPT system. The input/output characteristics of the IPT system are mainly determined by the coupling pad and the employed compensation topology. Verifying the operation of the coupling pad becomes challenging when the IPT application's required input/output characteristics exceed the experimental voltage range in laboratory environments. The same electrical stress is applied to the coupling pad through topology reconfiguration and resonance component tuning, and the input/output characteristics can be flexibly changed to present a guideline that can be tested in a laboratory environment. A 3-resonance component circuit allows for modeling various compensation topologies. The same electrical and heating stress are verified through a 3.3-kW experimental prototype.

As an advanced detection technique, the electrical yield meter (EYM) based on electrical resistance tomography has increasingly been applied in various industry fields. But EYM often works under high temperature differences, strong vibrations, high friction, and great corrosions, thus the faults are random and inevitable. How to effectively detect and find EYM faults remains unresolved, while the measurements of EYM failures/faults can provide a way to distinguish the faults themselves in practice. In this paper, an experimental platform is firstly built to simulate and construct the actual EYM working process, and thereby different types of EYM faults are reconstructed to obtain a set of sufficient normal and abnormal measurements. After performing the principal component analysis to capture the key features hidden in measurements, all faults are divided to six clusters by a typical c-means clustering algorithm. Therefore, a fault detection method is created along different measuring conditions and flowing patterns. Experimental evaluation in practical engineering shows that the proposed method can quickly and effectively distinguish fault types through actual measurements. Consequently, the proposed fault detection method is feasible and effective to ensure the relief and stable EYM measurement in practice.

Contrastive learning has revolutionized the field of computer vision, learning rich representations from unlabeled data, which generalize well to diverse vision tasks. Consequently, it has become increasingly important to explain these approaches and understand their inner workings mechanisms. Given that contrastive models are trained with interdependent and interacting inputs and aim to learn invariance through data augmentation, the existing methods for explaining single-image systems (e.g., image classification models) are inadequate as they fail to account for these factors and typically assume independent inputs. Additionally, there is a lack of evaluation metrics designed to assess pairs of explanations, and no analytical studies have been conducted to investigate the effectiveness of different techniques used to explaining contrastive learning. In this work, we design visual explanation methods that contribute towards understanding similarity learning tasks from pairs of images. We further adapt existing metrics, used to evaluate visual explanations of image classification systems, to suit pairs of explanations and evaluate our proposed methods with these metrics. Finally, we present a thorough analysis of visual explainability methods for contrastive learning, establish their correlation with downstream tasks and demonstrate the potential of our approaches to investigate their merits and drawbacks. [ABSTRACT FROM AUTHOR]

Dyslexia is a complex reading disorder characterized by difficulties in accurate or fluent word recognition, poor spelling, and decoding abilities. These challenges are not due to intellectual, visual, or auditory deficits. The diagnosis of dyslexia is further complicated by symptom variability, influenced by cultural and personal factors. This study leverages Virtual Reality (VR) advancements, eye movement tracking, and machine learning to create a virtual reading environment that captures eye movement data. This data extracts features such as eye movement metrics, word vectors, and saliency maps. We introduce a novel fusion model that integrates various machine learning algorithms to objectively and automatically assess dyslexia using physiological data derived from user interactions. Our findings suggest that this model significantly enhances the accuracy and efficiency of dyslexia diagnosis, marking an important advancement in educational technology and providing robust support for individuals with dyslexia. Although the sample size was limited to 10 dyslexic and 4 control participants, the results offer valuable insights and lay the groundwork for future studies with larger cohorts. [ABSTRACT FROM AUTHOR]

After a neurological injury, neurorehabilitation aims to restore sensorimotor function of patients. Technological assessments can provide high-quality data on a patient’s performance and support clinical decision making towards the most appropriate therapy. In this study, the ArmeoPower, a robotic exoskeleton for the upper extremities, was used to assess 12 neurological patients and 31 non-disabled participants performing various standardized single joint and frontal plane game-like exercises. From the collected data, kinematic metrics (End-Point Error, Hand-Path Ratio, reaction time, stability, Number of Velocity Peaks, peak, and mean Velocity) and the game score, were calculated and analyzed according to three criteria: the reliability (a), the difference between patients and non-disabled participants (b), as well as the influence of robotic movement assistance (c). In total, 39 metrics were analyzed and the following five most promising assessment variables for different exercises could be identified based on the three above-mentioned criteria: smoothness (RainMug (wrist)), mean speed (RainMug (wrist)), reaction time (Goalkeeper), maximum speed (HighFlyer (elbow)) and accuracy (Connect the dots), with the former showing good validity (rho=0.82, p=0.02) when comparing to the patient’s severity level. The results demonstrate feasibility to extract and analyze various kinematic metrics from the ArmeoPower, which can provide quantitative information about human performance during training and therapy. The generated data increases the understanding of the patient’s movement and can be used in the future in clinical research for better performance evaluation and providing more feedback options, leading towards a more personalized and patient-centric therapy. [ABSTRACT FROM AUTHOR]

Transcranial photobiomodulation (tPBM) has been widely studied for its potential to enhance cognitive functions of the elderly. However, its efficacy varies, with some individuals exhibiting no significant response to the treatment. Considering these inconsistencies, we introduce a machine learning approach aimed at distinguishing between individuals that respond and do not respond to tPBM treatment based on functional near-infrared spectroscopy (fNIRS) acquired before the treatment. We measured nine cognitive scores and recorded fNIRS data from 62 older adults with cognitive decline (43 experimental and 19 control subjects). The experimental group underwent tPBM intervention over a span of 12 weeks. Based on the comparison of the global cognitive score (GCS), merging the nine cognitive scores into a single representation, acquired before and after tPBM treatment, we classified all participants as responders or non-responders to tPBM with a threshold for the GCS change. The fNIRS data were recorded during the resting state, recognition memory task (RMT), Stroop task, and verbal fluency task. A regularized support vector machine was utilized to classify the responders and non-responders to tPBM. The most promising performance of our machine learning model was observed when using the fNIRS data collected during the RMT, which yielded an accuracy of 0.8537, an F1-score of 0.8421, sensitivity of 0.7619, and specificity of 0.95. To the best of our knowledge, this is the first study to demonstrate the feasibility of predicting the tPBM efficacy. Our approach is expected to contribute to more efficient treatment planning by excluding ineffective treatment options. [ABSTRACT FROM AUTHOR]

The calibration procedure for a wearable P300 brain-computer interface (BCI) greatly impact the user experience of the system. Each user needs to spend additional time establishing a decoder adapted to their own brainwaves. Therefore, achieving subject independent is an urgent issue for wearable P300 BCI needs to be addressed. A dataset of electroencephalogram (EEG) signals was constructed from 100 individuals by conducting a P300 speller task with a wearable EEG amplifier. A framework is proposed that initially improves cross- subject consistency of EEG features through a common feature extractor. Subsequently, a simple and compact convolutional neural network (CNN) architecture is employed to learn an embedding sub-space, where the mapped EEG features are maximally separated, while pursuing the minimum distance within the same class and the maximum distance between different classes. Finally, the model’s generalization capability was further optimized through fine-tuning. Results: The proposed method significantly boosts the average accuracy of wearable P300 BCI to $73.23\pm 7.62$ % without calibration and $78.75\pm 6.37$ % with fine-tuning. The results demonstrate the feasibility and excellent performance of our dataset and framework. A calibration-free wearable P300 BCI system is feasible, suggesting significant potential for practical applications of the wearable P300 BCI system. [ABSTRACT FROM AUTHOR]

Brain networks/graphs have been widely recognized as powerful and efficient tools for identifying neurological disorders. In recent years, various graph neural network models have been developed to automatically extract features from brain networks. However, a key limitation of these models is that the inputs, namely brain networks/graphs, are constructed using predefined statistical metrics (e.g., Pearson correlation) and are not learnable. The lack of learnability restricts the flexibility of these approaches. While statistically-specific brain networks can be highly effective in recognizing certain diseases, their performance may not exhibit robustness when applied to other types of brain disorders. To address this issue, we propose a novel module called Brain Structure Inference (termed BSI), which can be seamlessly integrated with multiple downstream tasks within a unified framework, enabling end-to-end training. It is highly flexible to learn the most beneficial underlying graph structures directly for specific downstream tasks. The proposed method achieves classification accuracies of 74.83% and 79.18% on two publicly available datasets, respectively. This suggests an improvement of at least 3% over the best-performing existing methods for both tasks. In addition to its excellent performance, the proposed method is highly interpretable, and the results are generally consistent with previous findings. [ABSTRACT FROM AUTHOR]

Facial palsy evaluation (FPE) aims to assess facial palsy severity of patients, which plays a vital role in facial functional treatment and rehabilitation. The traditional manners of FPE are based on subjective judgment by clinicians, which may ultimately depend on individual experience. Compared with subjective and manual evaluation, objective and automated evaluation using artificial intelligence (AI) has shown great promise in improving traditional manners and recently received significant attention. The motivation of this survey paper is mainly to provide a systemic review that would guide researchers in conducting their future research work and thus make automatic FPE applicable in real-life situations. In this survey, we comprehensively review the state-of-the-art development of AI-based FPE. First, we summarize the general pipeline of FPE systems with the related background introduction. Following this pipeline, we introduce the existing public databases and give the widely used objective evaluation metrics of FPE. In addition, the preprocessing methods in FPE are described. Then, we provide an overview of selected key publications from 2008 and summarize the state-of-the-art methods of FPE that are designed based on AI techniques. Finally, we extensively discuss the current research challenges faced by FPE and provide insights about potential future directions for advancing state-of-the-art research in this field. [ABSTRACT FROM AUTHOR]

Assessing the motor impairments of individuals with neurological disorders holds significant importance in clinical practice. Currently, these clinical assessments are time-intensive and depend on qualitative scales administered by trained healthcare professionals at the clinic. These evaluations provide only coarse snapshots of a person’s abilities, failing to track quantitatively the detail and minutiae of recovery over time. To overcome these limitations, we introduce a novel machine learning approach that can be administered anywhere including home. It leverages a spatial-temporal graph convolutional network (STGCN) to extract motion characteristics from pose data obtained from monocular video captured by portable devices like smartphones and tablets. We propose an end-to-end model, achieving an accuracy rate of approximately ${76}.{6}\%$ in assessing children with Cerebral Palsy (CP) using the Gross Motor Function Classification System (GMFCS). This represents a ${5}\%$ improvement in accuracy compared to the current state-of-the-art techniques and demonstrates strong agreement with professional assessments, as indicated by the weighted Cohen’s Kappa ($\kappa _{\textit {lw}} = {0}.{733}$). In addition, we introduce the use of metric learning through triplet loss and self-supervised training to better handle situations with a limited number of training samples and enable confidence estimation. Setting a confidence threshold at ${0}.{95}$ , we attain an impressive estimation accuracy of ${88}\%$. Notably, our method can be efficiently implemented on a wide range of mobile devices, providing real-time or near real-time results. [ABSTRACT FROM AUTHOR]

Brain-computer interfaces (BCIs) are anticipated to improve the efficacy of rehabilitation for people with motor disabilities. However, applying BCI in clinical practice is still a challenge due to the great diversity of patients. In the current study, a novel action observation (AO) based BCI was proposed and tested on stroke patients. Ten non-hemineglect patients and ten hemineglect patients were recruited. Four AO stimuli were designed, each presenting a decomposed action to complete the reach-and-grasp task. EEG data and eye movement data were collected. Eye movement data was utilized to analyze the reasons for individual differences in BCI performance. Task discriminative component analysis was utilized to perform online target detection. The results showed that the designed AO-based BCI could simultaneously induce steady state motion visual evoked potential (SSMVEP) from the occipital region and sensory motor rhythm from the sensorimotor region in stroke patients. The average online detection accuracy among the four AO stimuli reached 67% within 3 s in the non-hemineglect group, while the accuracy only reached 35% in the hemineglect group. Gaze metrics showed that the average total duration of fixations during the stimulus phase in the hemineglect group was only 1.31 s ± 0.532 s which was significantly lower than that in the non-hemineglect group. The results indicated that hemineglect patients have difficulty gazing at the AO stimulus, resulting in poor detection performance and weak desynchronization in the sensorimotor region. Furthermore, the degree of neglect is inversely proportional to the target detection accuracy in hemineglect stroke patients. In addition, the gaze metrics associated with cognitive load were significantly correlated with the accuracy in non-hemineglect patients. It indicated the cognitive load may affect the AO-based BCI. The current study will expedite the clinical application of AO-based BCI. [ABSTRACT FROM AUTHOR]

Muscle fatigue significantly impacts coordination, stability, and speed in daily activities. Accurate assessment of muscle fatigue is vital for effective exercise programs, injury prevention, and sports performance enhancement. Current methods mostly focus on individual muscles and strength evaluation, overlooking overall fatigue in multi-muscle movements. This study introduces a comprehensive muscle fatigue model using non-negative matrix factorization (NMF) weighting. NMF is employed to analyze the duration multi-muscle weight coefficient matrix (DMWCM) during synergistic movements, and four electromyographic (EMG) signal features in time, frequency, and complexity domains are selected. Particle Swarm Optimization (PSO) optimizes feature weights. The DMWCM and weighted features combine to calculate the Comprehensive Muscle Fatigue Index (CMFI) for multi-muscle synergistic movements. Experimental results show that CMFI correlates with perceived exertion (RPE) and Speed Dynamic Score (SDS), confirming its accuracy and real-time tracking in assessing multi-muscle synergistic movements. This model offers a more comprehensive approach to muscle fatigue assessment, with potential benefits for exercise training, injury prevention, and sports medicine. [ABSTRACT FROM AUTHOR]

Dynamic functional connectivity (FC) analyses have provided ample information on the disturbances of global functional brain organization in patients with schizophrenia. However, our understanding about the dynamics of local FC in never-treated first episode schizophrenia (FES) patients is still rudimentary. Dynamic Regional Phase Synchrony (DRePS), a newly developed dynamic local FC analysis method that could quantify the instantaneous phase synchronization in local spatial scale, overcomes the limitations of commonly used sliding-window methods. The current study performed a comprehensive examination on both the static and dynamic local FC alterations in FES patients (N = 74) from healthy controls (HCs, N = 41) with resting-state functional magnetic resonance imaging using DRePS, and compared the static local FC metrics derived from DRePS with those calculated from two commonly used regional homogeneity (ReHo) analysis methods that are defined based on Kendall’s coefficient of concordance (KCC-ReHo) and frequency coherence (Cohe-ReHo). Symptom severities of FES patients were assessed with a set of clinical scales. Cognitive functions of FES patients and HCs were assessed with the MATRICS consensus cognitive battery. Group-level analysis revealed that compared with HCs, FES patients exhibited increased static local FC in right superior, middle temporal gyri, hippocampus, parahippocampal gyrus, putamen, and bilateral caudate nucleus. Nonetheless, the dynamic local FC metrics did not show any significant differences between the two groups. The associations between all local FC metrics and clinical characteristics manifested scores were explored using a relevance vector machine. Results showed that the Global Assessment of Functioning score highest in past year and the Brief Visuospatial Memory Test-Revised task score were statistically significantly predicted by a combination of all static and dynamic features. The diagnostic abilities of different local FC metrics and their combinations were compared by the classification performance of linear support vector machine classifiers. Results showed that the inclusion of zero crossing ratio of DRePS, one of the dynamic local FC metrics, alongside static local FC metrics improved the classification accuracy compared to using static metrics alone. These results enrich our understanding of the neurocognitive mechanisms underlying schizophrenia, and demonstrate the potential of developing diagnostic biomarker for schizophrenia based on DRePS. [ABSTRACT FROM AUTHOR]

Resting state electroencephalography (rsEEG) is widely used to investigate intrinsic brain activity, with the potential for detecting neurophysiological abnormalities in clinical conditions from neurodegenerative disease to developmental disorders. When interpreting quantitative rsEEG changes, a key question is: how much deviation from a healthy normal brain state indicates a clinically significant change? Here, we build on the existing rsEEG variability literature by quantifying how this baseline rsEEG range can be attributed to common but underinvestigated sources of variability: experiment day, time of day, and pre-recording exercise level. We found that even within individuals, frequency band powers and entropy measures can vary by 7% (sample entropy and relative alpha power) to 28% (absolute delta power). Absolute and relative delta power increased significantly after running, while relative theta power decreased significantly. Relative beta and gamma power were significantly higher in the afternoon compared to morning trials. Sample entropy and alpha power were relatively consistent. The coefficients of variability we found are similar to some clinical rsEEG effect sizes identified in prior literature, bringing into question the clinical significance of these effect sizes. Furthermore, time of day and activity level accounted for more rsEEG variability than experiment day, indicating the potential to reduce variability by controlling for these factors in repeated-measures studies. [ABSTRACT FROM AUTHOR]

The partition of a long urban arterial is necessary for efficient and reliable traffic signal coordination. Previous studies have utilized fixed detector data or aggregated vehicle trajectory data to measure the correlation between adjacent intersections for an arterial partition. However, the appropriate partition of a long urban arterial remains a challenge because of the difficulty in quantifying the correlation between any pair of intersections instead of two adjacent intersections. The availability of spatiotemporal continuous individual vehicle trajectory data provides opportunities for a more accurate dynamic urban arterial partition by constructing a correlation matrix between intersections. In this article, a novel three-step framework is proposed for a dynamic urban arterial partition using the data collected by automatic license plate recognition devices, which are widely installed in China. First, the arterial traffic origin-destination (OD) flow between any pair of intersections along an arterial is extracted, and a similarity model is proposed to identify the spatiotemporal pattern of the arterial traffic OD flow matrixes. Second, a community aggregation-based algorithm is developed to produce candidate arterial partition solutions using the distributed characteristics of the arterial traffic OD flow matrixes. Third, two new metrics are designed to determine the optimal arterial partitions considering the splits of vehicle trajectories and their coordination utilization in partitioned intersection groups. The proposed framework is implemented on an actual arterial in Kunshan, China. The comparison to conventional adjacent correlation methods indicates that the OD flow-based partition method can generate better signal coordination control performance, in terms of average delay, average speed, and so on. [ABSTRACT FROM AUTHOR]

How to drive safely in complex real-world traffic settings has long been a question and challenge for autonomous vehicles (AVs). Decision-making systems (DecSs) are the core of AVs, and their safety and rationality are crucial. Several decision-making techniques and algorithms have been applied to AVs; however, they are still subject to limitations and deficiencies, making it impossible to fully guarantee safety. Thus, it is necessary to conduct a safety assessment of the state performance of the DecSs of AVs to reduce driving risks and implement immediate safety measures. This article reviews relevant research on safety assessment methods of AVs’ DecSs. The insufficiency of the current DecSs in terms of safety is analyzed, and the importance and definition of decision-making safety assessment (DMSA) are established. On this basis, assessment indexes related to DMSA are analyzed. From the perspectives of methodology and industrialization development, the existing safety assessment methods are summarized categorically, their applicability and limitations are discussed, and corresponding constructive strategies for improvement are proposed. Additionally, the deficiencies of the existing DMSA methods and the areas that need improvement are analyzed. In this regard, the future challenges and development opportunities of DMSA are proposed. This article provides a useful reference for the perfect DMSA method. [ABSTRACT FROM AUTHOR]

Algorithmic Decision-Making Systems (ADMS) fairness issues have been well highlighted over the past decade , including some facial recognition systems struggling to identify people of color. In 2021, Uber drivers filed a claim with the U.K. ’s employment tribunal for unfair dismissal resulting from automated facial recognition technology by Microsoft. Bias mitigation methods have been developed to reduce discrimination from ADMS. These typically operationalize fairness notions as fairness metrics to minimize discrimination. We refer to ADMS to which bias mitigation methods have been applied as “mitigated ADMS” or, in the singular, a “mitigated system.” [ABSTRACT FROM AUTHOR]

Physical unclonable functions (PUFs) are a handy security primitive for resource-constrained devices. They offer an alternative to the resource-intensive classical hash algorithms. Using the IC differences resulting from the fabrication process, PUFs give device-specific outputs (responses) when given the same inputs (challenges). Hence, without using a device-specific key, PUFs can generate device-specific responses. FPGAs are one of the platforms that are heavily studied as a candidate for PUF implementation. The idea is that a PUF that is designed as an HDL code can be used as part of the static design or as a dynamic accelerator. Previous works studied PUF implementation as part of the static design. In contrast to the state-of-the-art, this letter studies PUFs when used as runtime reconfigurable accelerators. In this letter, we find that not all regions of an FPGA are equally suitable for implementing different PUF types. Regions, where clock routing resources exist, are the worst suited for PUF implementation. Moreover, we find out that for certain PUF types, the property of dynamic partial reconfiguration can lead to performance degradation if not applied carefully. When static routing passing through the region increases, the PUF performance degrades significantly. [ABSTRACT FROM AUTHOR]

Point cloud registration, which effectively coincides the source and target point clouds, is generally implemented by geometric metrics or feature metrics. In terms of resistance to noise and outliers, feature-metric registration has less error than the traditional point-to-point corresponding geometric metric, and point cloud reconstruction can generate and reveal more potential information during the recovery process, which can further optimize the registration process. In this paper, CFNet, a correspondence-free point cloud registration framework based on feature metrics and reconstruction metrics, is proposed to learn adaptive representations, with an emphasis on optimizing the network. Considering the correlations among the paired point clouds in the registration, a feature interaction module that can perceive and strengthen the information association between point clouds in multiple stages is proposed. To clarify the fact that rotation and translation are essentially uncorrelated, they are considered different solution spaces, and the interactive features are divided into two parts to produce a dual branch regression. In addition, CFNet with its comprehensive objectives estimates the transformation matrix between two input point clouds by minimizing multiple loss metrics. The extensive experiments conducted on both synthetic and real-world datasets show that our method outperforms the existing registration methods. [ABSTRACT FROM AUTHOR]

During group interactions, we react and modulate our emotions and behaviour to the group through phenomena including emotion contagion and physiological synchrony. Previous work on emotion recognition through video/image has shown that group context information improves the classification performance. However, when using physiological data, literature mostly focuses on intrapersonal models that leave-out group information, while interpersonal models are unexplored. This paper introduces a new interpersonal Weighted Group Synchrony approach, which relies on Electrodermal Activity (EDA) and Heart-Rate Variability (HRV). We perform an analysis of synchrony metrics applied across diverse data representations (EDA and HRV morphology and features, recurrence plot, spectrogram), to identify which metrics and modalities better characterise physiological synchrony for emotion recognition. We explored two datasets (AMIGOS and K-EmoCon), covering different group sizes (4 vs dyad) and group-based activities (video-watching vs conversation). The experimental results show that integrating group information improves arousal and valence classification, across all datasets, with the exception of K-EmoCon on valence. The proposed method was able to attain mean M-F1 of $\approx$ ≈ 72.15% arousal and 81.16% valence for AMIGOS, and M-F1 of $\approx$ ≈ 52.63% arousal, 65.09% valence for K-EmoCon, surpassing previous work results for K-EmoCon on arousal, and providing a new baseline on AMIGOS for long-videos. [ABSTRACT FROM AUTHOR]

Two people's physiological responses become more similar as those people talk or cooperate, a phenomenon called physiological synchrony. The degree of synchrony correlates with conversation engagement and cooperation quality, and could thus be used to characterize interpersonal interaction. In this study, we used a combination of physiological synchrony metrics and pattern recognition algorithms to automatically classify four different dyadic conversation scenarios: two-sided positive conversation, two-sided negative conversation, and two one-sided scenarios. Heart rate, skin conductance, respiration and peripheral skin temperature were measured from 16 dyads in all four scenarios, and individual as well as synchrony features were extracted from them. A two-stage classifier based on stepwise feature selection and linear discriminant analysis achieved a four-class classification accuracy of 75.0% in leave-dyad-out crossvalidation. Removing synchrony features reduced accuracy to 65.6%, indicating that synchrony is informative. In the future, such classification algorithms may be used to, e.g., provide real-time feedback about conversation mood to participants, with applications in areas such as mental health counseling and education. The approach may also generalize to group scenarios and adjacent areas such as cooperation and competition. [ABSTRACT FROM AUTHOR]

This paper presents a novel survey of high level synthesis (HLS) based hardware security approaches for reusable intellectual property (IP) cores used in consumer electronics and computing systems. A succinct review of all major HLS based hardware security approaches applied on reusable IP cores, along with their design flow and security analysis, is provided. The paper presents a detailed design flow of hardware integrated circuits (ICs) along with vulnerability points where potential attacks/threats are possible. Trustworthy and untrustworthy regimes in the design flow have also been highlighted in the discussion. Further, a discussion of detective and preventive control-based HLS hardware security approaches used for hardware IP cores has also been presented, including an analysis of prominent structural obfuscation, logic locking (logic encryption), and IP core protection (IPP) techniques. Each approach has been lucidly explained in terms of its threat model, algorithm, and security analysis. Finally, a security comparison of hardware IP obfuscation approaches in terms of strength of obfuscation security metric as well as a security comparison of IPP approaches in terms of probability of coincidence security metric, have also been introduced. [ABSTRACT FROM AUTHOR]

Approximate computing is an evolving paradigm that aims to improve the power, speed, and area in neural network applications that can tolerate errors up to a specific limit. This letter proposes a new multiplier architecture based on the algorithm that adapts the approximate compressor from the existing and proposed compressors’ set to reduce error in the respective partial product columns. Further, the error due to the approximation in the proposed multiplier is corrected using a simple error-correcting module. Results prove that the power and power–delay product (PDP) of an 8-bit multiplier is improved by up to 39.9% and 43.6% compared with the exact multiplier and 27.5% and 23.9% compared to similar previous designs. The proposed multiplier is validated using image processing and neural network applications to prove its efficacy. [ABSTRACT FROM AUTHOR]

The representation of temperature measurement methods is worldwide, as is the number of industries that use them. Their utilization allows one to assess the object’s temperature relatively accurately and further according to the defined dependencies to evaluate its condition. In the medical sector, non-invasive temperature measurement methods are increasingly gaining ground. In our work, we focus on the design of a system for measuring skin temperature using digital DS18B20 thermometers. The technical design consists of the mechanical design of the sensor housing based on the requirements of healthcare professionals, the design of a 1-wire bus active pull-up driver with improved features, and the user software, allowing for the future implementation in various, not necessarily medical environments. The main contribution lies in optimizing the hardware design concerning the components’ parameters and surge protection. In the medical field, the aim was to compare changes in skin temperature in the shoulder region depending on the type of muscle contraction during static and dynamic components of movement and to record the thermal work of the muscle during activity in real-time. We experimented by applying temperature sensors to the deltoid muscle of the dominant limb in a temperature environment of 25.76 °C. In the experimental measurements, we demonstrated that during the three selected muscle work movements, there was a temperature change in the respective regions of the deltoid muscle for each of the three, which the temperature sensors could capture. Our results were statistically significant when comparing the baseline values, and the highest temperature reached. The level of significance $\alpha \leqslant 0.05$ and a low percentage of errors were valid.

The increased availability and affordability of 3D scanning technology due to recent advancements have made pre and post-operative analysis for facial surgeries more convenient and accurate through the use of 3D patient scans. However, to perform a more comprehensive analysis using area and volume measurements, it is necessary to develop new software tools that include the appropriate algorithms. Without the right algorithms, any software tool cannot perform its intended tasks effectively. In this study, we introduce two algorithms along with their open-source code to calculate the surface area and volume of a specific subsection of a 3D model. The first algorithm partitions the area/volume into smaller segmentations and applies raycasting to find the points on the face surface, and then finds an approximate measurement depending on the resolution selected. In contrast, the second algorithm utilizes the vertex information stored in the 3D model file to calculate the area and volume measurements for the designated region. The advantage of the second algorithm is that it examines every point in the model, resulting in the highest degree of precision possible. We have also used ChatGPT to generate the code for area/volume calculation and found that the logic of the ChatGPT-generated code is surprisingly similar to the second algorithm. We have presented the results of the measurements on certain regions of the face for both of the algorithms and compared the temporal performance of the algorithms.

Open government data (OGD) is an e-governance practice that aims to increase public resource efficiency and improve service delivery for citizens. Saudi Arabia launched its first open data initiative in 2011, with the goal of maximizing the economic impact of open data locally. This empirical study aimed to assess the effects of the Saudi OGD initiative by conducting a literature review and monitoring the impact of Saudi state-of-the-art open data. After observation and analysis, a systematic Monitor, Analysis, Action, and Review (MAAR) approach was followed to conduct a proposed solution that addresses quality shortcomings that consequently affect open data reusability. The methodology for creating the proposed solution included selecting, formulating, and weighing the quality characteristics that define a valuable OGD. The proposed solution is a framework for assessing the OGD quality using metrics that are compatible with the Saudi Open Data Portal (od.data.gov.sa). To review this proposed quality assessment framework and test whether it achieved its predicted outcomes, the results of the proposed solution were evaluated using local data samples from the Saudi Open Data Portal.