Visual grounding of remote sensing (RSVG) aims to detect the specific objects associated with query expressions in remote sensing (RS) images.

Pedestrian and vehicle trespassing is a significant problem in the railroad industry. According to the Federal Railroad Administration, 94% of rail-related deaths occur at railway crossings due to trespassing. These circumstances often happen due to individuals illegally taking shortcuts in off-limits rail zones or ignoring traffic signals at grade crossings. This is especially a problem present in NJ Transit rail lines, with there being reports every year of pedestrians or vehicles being struck by a train due to trespassing. NJ Transit collects a large amount of surveillance data across many of their rail crossings, but it is very expensive and inefficient to manually review this footage. This study proposes a preliminary implementation of a YOLO convolutional neural network to detect railway trespassers in video footage, creating anchor boxes around the train, pedestrians, vehicles, and railroads in the input data. A model was created that improves upon previous models through the use of a single-stage for detection that improves speed and computational resources. This neural network, constructed using libraries from Darknet and YOLOv7, was found to be extremely accurate in its detections and has been tested in a wide variety of situations. The data collected from this can be used to provide NJ Transit necessary data regarding the types of crossings that are most dangerous for pedestrians, the times of day that trespassing occurs, and the typical causes of trespass. Future studies within this topic may involve a live alert-based system for trespass and the implementation of specific regions of interest the model may focus on.

Rising shares of renewable generation raises uncertainty and thus the number of possible power flow scenarios in the power system, which in turn increases the possibility for unforeseen contingencies, such as power line or generator failures and their combinations. Therefore, operators cannot longer rely only on operational experience to deal with every contingency. Our proposed method involves identifying the most effective countermeasures to minimize the impact of contingencies on the power system. We take into account various options, such as load shedding, adjusting phase-shifting transformer angles, and injecting active power using fast elements. The proposed approach considers primary control of generators and its limitations in order to compensate for power imbalances in the system. The problem is formulated as a mixed-integer linear optimization problem, employing DC power flow equations. The applicability of the approach is evaluated on the IEEE 39-bus system, and the scalability of the approach is shown on five systems with up to 6470 buses.

Objective: Selective auditory attention decoding (AAD) algorithms process brain data such as electroencephalography to decode to which of multiple competing sound sources a person attends. Example use cases are neuro-steered hearing aids or communication via brain-computer interfaces (BCI). Recently, it has been shown that it is possible to train such AAD decoders based on stimulus reconstruction in an unsupervised setting, where no ground truth is available regarding which sound source is attended. In many practical scenarios, such ground-truth labels are absent, making it, moreover, difficult to quantify the accuracy of the decoders. In this paper, we aim to develop a completely unsupervised algorithm to estimate the accuracy of correlation-based AAD algorithms during a competing talker listening task.Methods: We use principles of digital communications by modeling the AAD decision system as a binary phase-shift keying channel with additive white gaussian noise. Results: We show that the proposed unsupervised performance estimation technique can accurately determine the AAD accuracy in a transparent-for-the-user way, for different amounts of training and estimation data and decision window lengths. Furthermore, since different applications demand different targeted accuracies, our approach can estimate the minimal amount of training required for any given target accuracy.Conclusion: Our proposed estimation technique accurately predicts the performance of a correlation-based AAD algorithm without access to ground-truth labels. Significance: In neuro-steered hearing aids, the accuracy estimates provided by our approach could support time-adaptive decoding, dynamic gain control, and neurofeedback. In BCIs, it could support a robust communication paradigm with accuracy feedback for caregivers.

The electrocardiogram (ECG) serves as a valuable diagnostic tool, providing crucial information about life-threatening cardiac conditions such as Atrial Fibrillation and Myocardial Infarction. A prompt and efficient assessment of ECG exams in environments like emergency rooms (ERs) can significantly improve the chances of survival for high-risk patients. In this study, we have developed an artificial intelligence-driven screening system specifically designed to analyze 12-lead ECG images. Our proposed method has been trained on an extensive dataset comprising 99,746 12-lead ECG exams collected from the ambulatory section of a tertiary hospital. The primary objective was to accurately classify the exams into three classes: Normal (N), Atrial Fibrillation (AFib), and Other (O). The evaluation of our method resulted in AUROC scores of 95.3%, 99.1%, and 93.3% for N, AFib, and O, respectively. To further validate our approach, we conducted evaluations using the Chinese Physiological Signal Challenge database. In this evaluation, we achieved AUROC scores of 91.8%, 97.5%, and 70.4% for the classes N, AFib, and O, respectively. Additionally, we assessed our method using 1,074 exams acquired in the ER, and achieved AUROC values of 98.3%, 98.0%, and 97.7% for the classes N, AFib, and O, respectively. Finally, we developed and deployed a system with a trained model within the ER of a tertiary hospital for research purposes. The system automatically retrieves newly captured ECG chart images from the Picture Archiving and Communication System (PACS) within the ER. These images undergo necessary preprocessing steps and serve as input for our proposed classification method. This comprehensive approach has resulted in the establishment of an efficient and versatile end-to-end framework for ECG classification. The results of our study highlight the potential of leveraging artificial intelligence in the screening of ECG exams, offering a promising solution for the rapid assessment and prioritization of patients in the ER.

World applications could be optimized using AI approaches. However, due to the ever-increasing complexity of operations and services, it is challenging to address today's industrial systems' heterogeneous requirements for high reliability, ultra-low latency, high data rates, and limited resources through conventional neural networks. In this study, we suggest an AI-based modular architecture for handling an industrial human-made system, driven by Blockchain and 5G network slicing services. In order to successfully tackle the targeted issue by dividing it into subsystems based on AI and other methodologies, we aim to investigate a federated modular strategy powered by an intelligent storage system. The intelligent system removes human mistakes and uses AI to automate the system. This improves efficiency, and generalization capabilities regarding architecture and data. We examine a use case in the healthcare industry that has effectively demonstrated the advantages of our study.

This paper develops a scalable dc reconfigurable battery pack (RBP) and the accompanying hierarchical balancing algorithm. The battery cells in the RBP are divided into several modules, and each module is equipped with a slave controller that is only responsible for a subset of cells. The system controller performs the closed-loop control and allocates the desired voltage steps into battery modules. The hierarchical algorithm significantly reduces communication and computation requirements for large-scale applications. Moreover, considering the second-harmonic current (SHC) widely existing in a single-phase dc/ac system, which may induce the aging of battery cells, the dc RBP is used to suppress the SHC for the first time without an additional central dc/dc converter. Due to the high output voltage quality and low-frequency operation of the dc RBP, the filter size can be significantly reduced and the efficiency of the system can be improved. Experimental validation of the proposed control scheme is presented using a dc RBP containing 48 3-Ah 18650 lithium-ion cells. Experimental results show that the proposed method can effectively balance battery cells and suppress the SHC.

Soil moisture can vary spatially at the scale of agricultural fields (~10 -100 m), which is generally too fine to resolve using passive radiometric methods. Active radar provides an opportunity for finer resolution measurements; in particular, the interferometric synthetic aperture radar (InSAR) closure phase parameter is sensitive to changing soil moisture. We have developed a model showing that systematic non-zero closure phase can result from scattering from objects at different depths in a medium of time-varying dielectric, such as from changes in soil moisture. The model predicts that interference between surface and subsurface reflections are needed for closure phase to be non-zero. We find that, under certain circumstances, we can estimate soil moisture from closure phase using a data reduction approach that includes a cumulative sum of closure phase over time and removal of a trend. The correlation between cumulative closure phase and soil moisture suggests that the closure phase is related to the change in soil moisture. We examined a large test region in Oklahoma, where the detrended cumulative closure phase from Sentinel-1 data demonstrates some agreement with in situ soil moisture measurements. In other areas, the match is weaker, implying a terrain dependence for the quality of fit. Cumulative InSAR closure phase promises to provide a valuable new method to remotely estimate soil moisture.

Conceptual scaling is a method, developed in the framework of Formal Concept Analysis (FCA), that transforms a many-valued context (a collection of objects, described by attributes with values, such as name, size or color) into a formal context (a binary schema from which formal concepts are derived). Relational scaling extends conceptual scaling, by taking relations between objects into account. Previous work applies relational scaling to a relational database, transforming the relational database into a relational structure. Relational structures play the role of formal contexts in a relational variant of FCA. In this paper, we present a similar approach, which applies relational scaling to RDFS ontologies.

This study showcases Convolutional Neural Networks’ (CNNs) potential in detecting breast cancer from ultrasound images. Impressive accuracy highlights their value in precise and efficient diagnosis, paving the way for further research. The CNN’s remarkable accuracy, even with a small dataset of 1,500 images, suggests enhanced performance with larger and diverse datasets. CNNs for breast cancer detection via ultrasound offer a promising, potent approach to revolutionize medical imaging and diagnosis, boosting accuracy and reliability.

This paper introduces a framework aimed at improving the lifestyle of individuals engaged in health self-management programs by providing them with valuable and actionable insights derived from their personal health data. The framework is designed to autonomously discover, optimize, and deliver these insights to the participants. To evaluate the effectiveness of the framework, an experiment was conducted where participants were provided with insights related to their sleep and physical activity. The results demonstrate that the proposed framework, which incorporates feedback-driven optimization, effectively recommends insights that align closely with the behavior and preferences of the participants. These findings highlight the potential of the framework to enhance the impact of health self-management programs.

In four participants with upper limb loss, we characterized the quality and location of sensation elicited through electrical stimulation of regenerative peripheral nerve interfaces (RPNIs) over time (up to 54 months post implantation of wires). We also measured sensory perception and discomfort thresholds, sensitivity to changes in stimulation amplitude, and ability to differentiate objects of different stiffness and sizes.

Knowledge of fuzzy concepts defined on real-valued domains may be imprecise due to differing interpretations of the concept as well as to respondents’ lack of confidence in their own judgements.  This second source of imprecision has largely been ignored.  We describe a novel procedure to form an interval type-3 fuzzy set on the domain from two interval type-2 (IT2) fuzzy sets that model each of these sources of imprecision.  The first IT2 fuzzy set is on the real-valued domain and represents a consensus of opinions about the concept description.  The second IT2 fuzzy set is on the unit interval and represents a consensus of respondents’ sentiment about their own judgements.  These sets are either generated from interval inputs using the Hao-Mendel Approach, or from respondents’ selections from ordinal vocabularies whose words have IT2 representations as described by Hao and Mendel.  We show more generally how imprecision in hierarchical estimates of knowledge can be incorporated into type-n fuzzy set representations of concepts.  These representations belong to a class of general fuzzy sets constructed from IT2 fuzzy sets that support efficient type reduction.

This work presents a novel Deep Autoencoder neural network which is developed for removing EOG, EMG, and motion artifacts from single-channel EEG signal. The proposed algorithm has been demonstrated to have a desirable perfromance on EEG artifact removal. More importantly, this model has been also implemented into Android smartphone fo the first time via TensorFlow Lite Library, with a fast signal processing speed after enabling hardware/software acceleration on smartphones, which outperforms the gold standard ICA algorithm from the perspectives of computation speed and power consumption, showing promising applications in future mobile EEG and Brain-Computer Interfaces.

UPS Industrial Engineering Training & Development team consists of training managers responsible for hiring, orienting, and educating interns and new employees. Current procedures involve storing overlapping and inconsistent data in various Excel files, such as OneNote, Sharepoint, and local workspaces. This can lead to human mistakes and challenges in maintaining accurate historical and present data. To streamline the hiring process, the Industrial Engineering Team Dashboard (IETD) was developed as a web tool for training managers. The IETD streamlines the screening process, candidate tracking, and data entry, reducing repetition and data loss associated with existing entry processes. Multiple dashboards will be used to implement data visualization, focusing on specific aspects of the training manager’s function, such as retention rates, new-hire training programs, and intern recruitment. This will allow training managers to focus on candidate development and candidate retention, ultimately improving their overall efficiency and effectiveness.

This paper presents simulation and measurement results from a variety of compact PCBbased antenna designs suitable for wireless systems operating at 2.4 GHz using realistic implementation conditions in contrast to idealized, best-case, configurations. In particular, the input impedance, radiation pattern, and radiation efficiency were measured using a variety of simulated and physically constructed prototypes and test fixtures. The results were compared with published application notes and manufacturer specifications to obtain better estimates of antenna performance in implementations, such as centimeterscale wireless sensor nodes with size and/or weight constraints. Notable deviations from published expectations of performance were exhibited in both simulation and measurement, including absolute gain, input impedance, and radiation pattern. Reasonably high radiation efficiency was also demonstrated. Significant deviations from both simulation and manufacturer-published results were observed in all investigated parameters. Differences from datasheet values of as much as 11 dB in antenna match and 12 dBi in maximum gain are reported. These results indicate a number of important considerations for a wireless sensor system designer, including that chip antennas are highly preferable to PCB trace antennas if cost is not a concern, implementations of designs tend to decrease directivity, and as much as half the efficiency expected from simulation is seen in realistic implementation.

We examine the optimal role, or use case, for a space solar power system (SSPS) in an electrical grid by using a full year of historical load data from three U.S. cities in different climate regions to drive a simulation of load and supply. We include the impact of the eclipsing of the power station during each equinox season. SSPS operation through eclipse requires over-sizing the space power station relative to the load, and storing the difference in energy for use during the eclipse seasons. We found the optimal use case for SSPS in cities north (south) of the subtropic region is providing the base-load supply and that load chasing may be the optimal use case for cities in the sub-tropical and tropical regions. Further analysis is needed to confirm this latter result.

Visible-infrared person re-identification (VI-ReID) is a challenging cross-modality image retrieval task. Compared to visible modality person re-identification that handles only the intra-modality discrepancy, VI-ReID suffers from an additional modality gap. Most existing VI-ReID methods achieve promising accuracy in a supervised setting, but the high annotation cost limits their scalability to real-world scenarios. Although a few unsupervised VI-ReID methods already exist, they typically rely on intra-modality initialization and cross-modality instance selection, despite the additional computational time required for intra-modality initialization. In this paper, we study the fully unsupervised VI-ReID problem and propose a novel cross-modality hierarchical clustering and refinement (CHCR) method by promoting modality-invariant feature learning and improving the reliability of pseudo-labels. Unlike conventional VI-ReID methods, CHCR does not rely on any manual identity annotation and intra-modality initialization. First, we design a simple and effective cross-modality clustering baseline that clusters between modalities. Then, to provide sufficient inter-modality positive sample pairs for modality-invariant feature learning, we propose a cross-modality hierarchical clustering algorithm to promote the clustering of inter-modality positive samples into the same cluster. In addition, we develop an inter-channel pseudo-label refinement algorithm to eliminate unreliable pseudo-labels by checking the clustering results of three channels in the visible modality. Extensive experiments demonstrate that CHCR outperforms state-of-the-art unsupervised methods and achieves performance competitive with many supervised methods.