X-ray computed tomography (CT) is an important non-destructive imaging technique, particularly in clinical diagnostics. Even with the latest innovations like dual-energy and photon-counting CT, the image contrast is solely generated from attenuation in the tissue. An extension – fully compatible with these novelties – is dark-field CT, which retrieves an additional, so-called dark-field contrast. Unlike the attenuation channel, the dark-field channel is sensitive to tissue microstructure and porosity below the resolution of the imaging system, which allows additional insights into the health of the lung tissue or the structure of calcifications. The potential clinical value has been demonstrated in several preclinical studies and recently also in radiography patient studies. Just recently the first dark-field CT for the human body was established at the Technical University of Munich and in this paper, we discuss the performance of this prototype. We evaluate the interferometer components and the imposed challenges that the integration into the CT gantry brings by comparing the results to simulations and measurements at a laboratory setup. The influence of the clinical X-ray source on the Talbot-Lau interferometer and the impact of vibrations, which are immanent on the clinical CT gantry, are analyzed in detail to reveal their characteristic frequencies and origin. A beam hardening correction is introduced as an important step to adapt to the poly-chromatic spectrum and make quantitative dark-field imaging possible. We close with an analysis of the image resolution and the applied patient dose, and conclude that the performance is sufficient to suggest initial patient studies using the presented dark-field CT system.

Experiments are conducted on a series of remote sensing images (NWPU VHR-10). Detection & instance segmentation results demonstrate that our method provides better performance (significant improvement in accuracy) in terms of average precision (AP). The larger the value of AP is, the more accurate the prediction results and the better detection performance of the objects. Cascade Mask R-CNN framework with HRNet backbone for geospatial objects detection and instance segmentation from high-resolution remote sensing imagery.

Classical blockchain cryptographic primitives are susceptible to quantum computing technology due to its unprecedented growth because it has the potential to make current blockchain encryption techniques obsolete against quantum attacks. Shor algorithm can perform transactional hijacking and forge digital signatures to impersonate blockchain users. On the other hand, Grover algorithm can perform 51% attack and can accelerate brute force attack by detecting hash collisions to replace the blocks of a blockchain. With the advent of this threat, in attempt to mitigate the effects of quantum attacks on blockchain technology, the concept of post-quantum cryptography (PQC) or quantum resistant cryptography (QRC) has emerged. PQC involves cryptographic algorithms or cryptosystems that are robust against quantum attacks and are also applicable in blockchain for post quantum decentralization. In this article, we provide a brief overview of quantum threats to classical blockchain technology. We identify and discuss some important requirements for the implementation of PQC. Moreover, we also identify and explain notable unresolved challenges in implementing PQC algorithms. We also determine critical research questions for each identified challenge that are imperative to confer and requires discussion and interpretation by the researchers of this field.

In most stochastic computing designs, the Stochastic Number Generator (SNG) circuit is of primary importance, but it may require considerable area due to its complexity. This paper proposes a new design for random number generators (RNGs) to be used in SNGs. The proposed RNG design exploits the randomness between each bit of data to generate larger sets of random numbers by concatenating the modules of the proposed customized linear feedback shift registers (LFSRs). To efficiently generate random data, a plane of RNGs (consisting of multiple modules) is proposed; a so-called sliding window approach is utilized for reading data in both directions; (horizontal and vertical). Therefore, sets of random numbers are generated by doubling these datasets and inverting the duplicated datasets. Flip-Flops (FFs) are then used to isolate the datasets and reduce the correlation among them. Compared to existing designs found in the technical literature, the proposed RNG design offers many advantages such as small area per RNG, low power operation, large number of datasets, flexibility and higher accuracy. When compared to the LFSR plane, the proposed RNG design improve power dissipation by 47.03%, area per RNG by 48.71% and the number of datasets by 1466%, the size of each random by 183,000% and the autocorrelation by 98.19%. The proposed design has been used in a stochastic multiplier and adder; for these circuits, the mean square error is reduced by 40.62% for the multiplier circuit, and 26.79% for the adder circuit.

Blockchain systems rely on miners to produce blocks, who solely build blocks for their financial benefit. Within the competitive market system, only correct blocks get accepted, which ensures security of the network. As financial applications migrate into the blockchain, miners in this privileged position have identified ways to extract additional revenues (known as MEV) that affect users and  the overall security of the system. A possible direction for current research to resolve this exploitation is to separate the sequencing (building) and execution (mining/proposing) of transactions (PBS model). However, a weaker point of this model is the trust assumption between participants and the incentive model. This paper proposes a concept of separate incentive schemes for the work various participants conduct in their own interests.

This communication presents a concept for a novel means of augmenting noncooperative active radar maritime surveillance of ship targets over very wide areas of the earth’s surface. The concept employs configurations of small satellites (smallsats) deployed in geostationary orbit (GEO) in combination with multiple-input multiple-output (MIMO) beamforming in order to provide persistent (uninterrupted) surveillance of large, medium-sized, and, to various extents, small-sized vessels. An example is presented which would provide L-band surveillance of seas adjoining southern Europe, including the Mediterranean, Adriatic, Aegean, and Black Seas. An investigation of ship detection in the presence of sea clutter is included. This is followed by a more-general discussion of target classification, including methods of rejecting targets not lying on the sea surface. Data processing burdens are estimated and found to be substantial. Comparisons are made with alternative approaches. Lastly, in view of the large scale of such a prospective deployment, a risk reduction plan is offered, along with suggesting less-ambitious designs.

This paper proposes new designs for Reduced Precision Redundancy (RPR) systems using Approximation (RPAs). Reduced redundancy is accomplished by utilizing approximate modules, hence requiring substantially different designs for the decision hardware for generating an output. Some of proposed schemes deal with a single erroneous data word generated by a module (in the presence of single and multiple bit errors) using three modules as inputs to the decision hardware of the RPA (denoted as 2E1A and 1E2A where E (A) stands for exact (approximate)). The remaining RPA schemes operate under single and multiple bit errors in at most two modules out of four input modules (denoted as 2E2A). Different from RPRs found in the technical literature, the proposed RPAs operate using only logic operations (so no arithmetic unit is involved as decision hardware). The probability of providing an exact data word at the output of the RPA under the above error conditions is analytically found; simulation results are also provided. It is shown that the difference between simulated and analytical probabilities is at most 5%. Circuit based metrics (such as delay, power dissipation, and area) of the proposed designs are simulated and compared with RPR; the proposed designs outperform RPR in all metrics, except the area of the four-input RPA (2E2A) due to the larger number of modules. The application of the proposed RPAs to image processing is also presented; the results show that the proposed schemes are very efficient for such application.

In this paper it is shown that the mathematical framework that correctly describes the combined effects of polarization mode dispersion and polarization dependent losses is the irreducible spinor representation of the extended Lorentz Group. There are two kinds of states of polarizations that are relevant for the description of PMD-PDL effects. The optical process that allows to convert one kind into the other is identified as optical phase conjugation. The utility of the proposed mathematical framework is proven proposing a simple technique (based on optical phase conjugation) that cancels the PDL part of the impairments at the receiving end of the optical system.

Thermal management in 3D integrated circuits is a critical challenge due to their high computational density. Heat dissipation paths from top circuit layers through bottom layers to substrate are heavily constraining heat extraction. Various thermal management frameworks have been proposed to address thermal issues in different granularities. All these frameworks require a thermal evaluation stage that characterizes the thermal profile of large designs with fast runtime. In this work, we present a machine learning based thermal evaluation method that predicts all standard cell temperatures based on features extracted from circuit CAD files. We have built thermal resistance networks for 10 benchmark circuits. We performed simulations to achieve the thermal data, and trained the thermal model with the data. The model is highly accurate and can identify all over-heated cells that need to be thermally-optimized. Runtime overhead is minimal. For a 435k-cell SPARC T2 core, the runtime for predicting all cell temperatures is as small as 3.12s, which is negligible compared to the runtime of other physical design stages.

Intelligent systems rely heavily on sensors and energy storage devices. Materials used in these devices are being upgraded using a variety of solid-state, chemical, and micro-machining techniques. Silicon, despite having an indirect bandgap, is being used in photonics due to technological advancements in device fabrication. Si photonics are in high demand due to their abundance and compatibility with CMOS. However, with reduced chip size, the thermal management is an important aspect which needs to be taken care. Multilayered Porous Silicon (PS) nanostructures were synthesized electrochemically. A 20-period Fabry Perrot Interference filter (FPIF) and Distributed Bragg Reflector (DBR) made from porous silicon showed optical entrapment beneficial for waveguides.

The performance characterization of communication systems assisted by large reconfigurable intelligent surfaces (RISs) significantly depends on the adopted models for the underlying channels. Under unrealistic channel models,  the system performance may be over- or under-estimated which yields inaccurate conclusions for the system design. In this paper, we review five channel models that are chosen to progressively improve the modeling accuracy for large RISs. For each channel model, we highlight the underlying assumptions, its advantages, and its limitations. We  compare the system performance under the aforementioned channel models using  RIS configuration algorithms from the literature and a new scalable algorithm proposed in this paper specifically for the configuration of extremely large RISs.

Nonlinear optical materials, such as transparent conductive oxides, have recently drawn a lot of attention when being integrated into metasurfaces and allowing full-optical control of the surface response. Although several methods for modeling the nonlinear materials have been proposed in the literature, most of them have the limitations on being non-dispersive and of instantaneous response. In this paper, we present a straightforward integration of an extended Drude-Lorentz model that captures the local intensity response of nonlinear materials while being dispersive and allowing for inertial response via a low-pass filtering process. This method is integrated into standard finite difference time-domain (FDTD) implementation of Maxwell’s equations and the auxiliary differential equations approach of the Drude-Lorentz model is extended via local intensity-dependent parameters. A numerical demonstration shows the response for a thin film of nonlinear material, where the parameters across the sample are time-varying with respect to the local intensity of the fields. Therefore, showing a direct feedback of the field profile to the nonlinear response of the material, which is critical when incorporating such films in resonating meta-atoms.

In the manuscript, as a result of comparing experimental and calculated kerrograms, it is shown that during electro-optical measurements in nitrobenzene using the Kerr effect, a strong electric field and absorption of radiation from a probing laser affect the properties of the medium under study. An estimate of such influence on the value of the Kerr constant is made. It is found that in the tip-plane system when using a ruby laser, the Kerr constant near the tip changes. The established dependence of the Kerr constant on the distance to the tip made it possible to obtain a calculated kerrogram that coincides with the experimental one. Thus, it is shown that the measuring instruments can influence the results obtained, that is, under certain conditions, the electro-optical research method is invasive. This fact should be taken into account when processing experimental kerrograms.

In this paper, a mathematical approach was used to improve stereo vision. The speed at which objects in multiple stereo images could be matched was improved using keypoints, and the use of multiple stereo images helped to improve the accuracy of the system. A three camera stereo vision system was built as a testbed to validate the approaches proposed in this paper. The research in this project can be applied as a fast and less costly method to optimize stereo vision systems with many uses such as in household robotics and autonomous vehicles. Stereo vision is the process of passively recovering object depth from camera images by comparing two images from different cameras of the same scene. Distance to the object is computed by comparing the shift of an object between the two images. The objects that are closer in the scene will have a larger shift between the images, while objects that are further away will have a smaller shift. This shift of the objects in the image is known as the disparity. The larger the disparity, the closer the object. Sensitivity analysis was conducted on the stereo vision calculation formula to analyze how the separation between two stereo cameras affected the system’s ability to accurately compute the distance to objects. The results showed that as the separation between two stereo cameras is increased, the accuracy and range for distance calculation also increases. A three camera system was built with cameras 3, 4 and 7 inches apart. Current approaches for calculating object distance compares the stereo images by matching a block of pixels in one image to a corresponding block in the other image to identify the same object in both images. This method, known as “block matching”, is time consuming, preventing usability on autonomous vehicles that rely on real-time information. SIFT keypoint feature detection along with k-Nearest Neighbor (knn) matching were studied and used to match the objects between two images. Keypoints are points in an image with unique features that can be identified on a corresponding stereo image such as corners or edges. Each keypoint has its own descriptor, which is a mathematical array and can be compared to other keypoints in the corresponding stereo image in order to match them. This approach is faster because it only requires matching of important regions of the image. Using the 3 camera test bed, data was collected in two environments. First in a noisy environment with a lot of background objects. Second, with the same objects and  a plain  background.The use of keypoints was faster and provided real-time data that was used to filter out the matches between these keypoints led to more accurate results. Furthermore, cameras with higher separation between them (7 inches) were able to more accurately detect the distances to objects further away while cameras with 3 inches and 4 inches of separation could only find the distances to nearby objects accurately. Results showed that in the more noisy environments, the 3in and 4in camera separation system exceeded an error of ±5in when an object was 85in away, while the 7in camera separation system exceeded an error of ±5in when an object was 165in away. In less noisy environments, the 3in camera separation system exceeded an error of ±5in when object was 90in away, the 4in camera separation system exceeded an error of ±5in when object was 100in away, and the 7in camera separation system exceeded an error of ±5in when object was 200in away.

Internet of Everything (IoE) connects billions of people, devices, objects and things to the Internet for autonomous services provisioning. IoE benefits from the value created by the compound impacts of connecting people, processes, things, and data. The interoperability among heterogeneous components facilitates seamless communication exchange for service provisioning in almost all the fields of life. Apart from providing immense benefits in autonomous services provisioning, it suffers from implementation issues, architectural considerations, and security and privacy vulnerabilities. There is a lack of surveys that comprehensively discuss all the aspects of IoE. Therefore, we design this survey to discuss key effective implementation of IoE. We discuss enabling technologies, architectural components, and expectations for an efficient realization of IoE. We discuss state-of-the-art use cases, synergies, and implementations of IoE that could be used as a road-map for and effective implementation of IoE-enabled systems. We highlight key security, privacy, and trust issues. We identify lethal attack scenarios and highlight countermeasures against these attacks. Finally, we discuss current challenges, their solutions, and future research directions. This survey provides key insights for implementing future IoE systems.

Recommendation systems, as a kind of information filtering system, can understand users' interests based on their personal data or historical behavior records, and are widely used in Web applications such as e-commerce, search, and streaming websites. Based on the fact that users with similar preferences may be interested in similar items, most existing recommendation methods mine this synergistic information through user-item interactions, called collaborative filtering. Recent interaction information can better reflect users' dynamic interests over time. Another group of work is called temporal-based recommendation, which takes into account temporal information to model the user's dynamic interests. A temporal recommendation system is based on the user's temporal interaction data as a context to predict which item the user is most likely to interact with next time. Markov chain-based recommendation models are a typical example of temporal sequential recommendation, which assume that the next interaction is only related to the previous one. Recently, much work has used recurrent neural networks to model previous behaviors and use hidden states to predict the next behavior. However, almost all temporal recommendation methods model user embeddings based only on their own temporal interaction history, ignoring the heterogeneous information widely available in recommender systems, such as product attributes, and encounter cold-start problems when data are sparse and users have fewer interactions. In this paper we proposed a model, the heterogeneous dynamic graph neural network. This network uses a static graph encoder to process the node representation of each heterogeneous graph at different time steps, which includes the node attribute information and edge information, to capture the changing node information in all time steps of the dynamic graph and obtain a valid node representation at each time step. The long and short-term memory model is then used to aggregate the temporal information between different time steps and to mine the connections and interdependencies of different types of nodes between different time steps. The goal of the heterogeneous dynamic graph neural network is to capture the changing node representations of the same nodes as well as different nodes at different time steps in the graph network, expecting to better perform the task of node classification in dynamic graphs. In this paper, we validate the model of heterogeneous dynamic graph neural networks from data, experimenting with node classification tasks on real graph structured data including social networks, e-commerce networks and business review networks, and show that heterogeneous dynamic graph neural networks outperform the latest representation learning methods for static and dynamic graphs.

The development of an ambient radio frequency (RF) powered Internet of Things (IoT) wireless sensor system is reported. The system integrates all major components including multiple antennas, rectifiers, a power management unit (PMU), a microcontroller unit (MCU) with RF transceiver module and sensing unit. A novelty of this applied engineering innovation is the design and use of multi-band dual-polarized multiple antennas to handle the low ambient RF power density found in the everyday environment. Our prototype has demonstrated that it is able to sense temperature, pressure and humidity and wirelessly transmit these signals to a central server with a duty cycle of approximately 15 minutes, when the background ambient RF signals are as low as −28.6 dBm at 950 MHz. The proposed system can address the important issue of energy supply in the development of IoT providing an alternative energy source to batteries or solar cells. Ambient RF powered IoT devices do not require periodic replacement of their batteries or to be placed visible regions which makes them suitable for wall, ceiling and floor applications. Our ambient RF powered IoT wireless sensor system provides concrete proof, through demonstration, that harnessing ambient RF energy for powering IoT sensor systems is feasible.

For several decades, the offline handwriting recognition problem has escaped a satisfactory solution. In the field of online recognition, researchers have had more successful performance, but the ability to extract dynamic information from static images has not been well explored yet. In this paper, we introduce a novel multi-lingual word handwriting recovery framework based on a convolutional denoising autoencoder with an attention model for pen up / down, velocity and temporal order recovery. The proposed framework consists of extracting robust features from a handwriting image using a stacked denoising autoencoder and an encoder Bidirectional Gated Recurrent Unit (BGRU) model. Then, the obtained vectors are decoded to produce an online script with dynamic characteristics using a BGRU with temporal attention. Evaluation is done on a Latin and Arabic Online and offline handwriting character / word databases and the proposed framework achieves high competitive results. To the best of our knowledge, this is the first work of its kind.