Data-driven soft sensors provide a potentially cost-effective and more accurate modeling approach to measure difficult-to-measure indices in industrial processes compared to mechanistic approaches. Artificial intelligence (AI) techniques, such as deep learning, have become a popular soft sensors modeling approach in the area of machine learning and big data. However, soft sensors models based deep learning potentially lead to complex model structures and excessive training time. In addition, industrial processes often rely on distributed control systems (DCS) characterized by resource constraints. Herein, guided by spatial geometric, a lightweight geometric constructive neural network, namely LightGCNet, is proposed, which utilizes compact angle constraint to assign the hidden parameters from dynamic intervals. At the same time, a node pool strategy and spatial geometric relationships are used to visualize and optimize the process of assigning hidden parameters, enhancing inter-pretability. In addition, the universal approximation property of LightGCNet is proved by spatial geometric analysis. Two versions algorithmic implementations of LightGCNet are presented in this article. Simulation results concerning both benchmark datasets and the ore grinding process indicate remarkable merits of LightGCNet in terms of small network size, fast learning speed, and sound generalization. Note to Practitioners-Motivated by the conflict between the high resource requirements of the traditional data-driven soft sensors models and the generally inadequate resource of DCS, this article proposes a lightweight approach that offers significant advantages in terms of small network size and fast learning speed. In this work, we visualize the black-box modeling process of data-driven soft sensors models based on knowledge of spatial geometry. Preliminary experiments suggest that this approach is feasible.

Harmonic retrieval techniques are the foundation of radio channel sounding, estimation and modeling. This paper introduces a Deep Learning approach for joint delay-and Doppler estimation from frequency and time samples of a radio channel transfer function. Our work estimates the two-dimensional parameters from a signal containing an unknown number of paths. Compared to existing deep learning-based methods, the signal parameters are not estimated via classification but in a quasi-grid-free manner. This alleviates the bias, spectral leakage, and ghost targets that grid-based approaches produce. The proposed architecture also reliably estimates the number of paths in the measurement. Hence, it jointly solves the model order selection and parameter estimation task. Additionally, we propose a multi-channel windowing of the data to increase the estimator's robustness. We also compare the performance to other harmonic retrieval methods and integrate it into an existing maximum likelihood estimator for efficient initialization of a gradient-based iteration.

Rapid advancements have been made in artificial intelligence applications recently, and generative models have prominently emerged as effective tools for domain transfer, image enhancement, and simulation. However, when dealing with large-scale gigapixel images, the use of traditional patch-based image aggregation methods introduces checkerboard or blocking artifacts, which compromises image quality and fidelity. Here we propose a parametric kernel that is specifically designed to target the underlying grid structure to mitigate these artifacts. The proposed parametric kernels are validated using three medical imaging modalities for three different generative model tasks, demonstrating improved visual fidelity and quantitative quality evaluation of the generated patch-aggregated images. The proposed method is versatile and compatible with various generative models, offering a robust framework for artifact reduction that can be seamlessly adjusted by modifying kernel parameters, and they can be directly applied and extended to other imaging modalities that employ large-scale images, such as astronomy and satellite imaging. The findings of this study have significant implications for medical imaging applications: by mitigating aggregation artifacts, our approach enhances the overall quality of medical images synthesized with generative models, which is crucial for accurate clinical assessment and subsequent image analysis. Furthermore, the proposed kernels provide a general formulation that can be extended to unpaired tasks, semantic segmentation, classification networks, and other large field-of-view imaging applications.

This article introduces a new approach to address the spectrum scarcity challenge in 6G networks by implementing the enhanced licensed shared access (ELSA) framework. Our proposed auction mechanism aims to ensure fairness in spectrum allocation to mobile network operators (MNOs) through a novel weighted auction called the fair Vickery-Clarke-Groves (FVCG) mechanism. Through comparison with traditional methods, the study demonstrates that the proposed auction method improves fairness significantly. We suggest using spectrum sensing and integrating UAV-based networks to enhance efficiency of the LSA system. This research employs two methods to solve the problem. We first propose a novel greedy algorithm, named market share-based weighted greedy algorithm (MSWGA) to achieve better fairness compared to the traditional auction methods and as the second approach, we exploit deep reinforcement learning (DRL) algorithms, to optimize the auction policy and demonstrate its superiority over other methods. Simulation results show that the deep deterministic policy gradient (DDPG) method performs superior to soft actor critic (SAC), MSWGA, and greedy methods. Moreover, a significant improvement is observed in fairness index compared to the traditional greedy auction methods. This improvement is as high as about 27% and 35% when deploying the MSWGA and DDPG methods, respectively.

A document by Sven Ochs. Click on the document to view its contents.

Full-duplex Cognitive Radio (FD-CR) technology has the potential to significantly improve the spectral efficiency of next-generation wireless systems. However, residual self-interference (RSI), unavoidable in FD systems, represents a colored noise, particularly at mm-wave frequencies. This fact affects blind Spectrum Sensing (SS) algorithms, preventing them from maintaining a constant false alarm rate (CFAR). This causes a power-throughput trade-off resulting in significant performance degradation. Detection in colored noise has been addressed in the literature through time-domain whitening aided by offline training. However, this method is ineffective in the presence of time-varying self-interference (SI). Our paper proposes an adaptive filter-based whitening approach to allow a blind SS to retain the CFAR property in mobile scenarios without the need for offline training. Focusing on low-complexity adaptive filtering, we analytically demonstrate that both the Least Mean Squares (LMS) and the Recursive Least Squares Lattice (RLSL) filters enable the sphericity test to achieve the CFAR property in typical FD-CR scenarios. We highlight the advantages of RLSL over LMS, including faster convergence, superior tracking, and modularity, making it suitable for effective implementation, as well as for efficient low-complexity SI cancellation. Numerical results confirm superior performance in high RSI power scenarios compared to offline solutions and LMS.

In safety-critical control for permanent magnet synchronous motors (PMSMs), overshooting after adding a spontaneous load is a crucial metric, leading to the unexpected motion of driving equipment, which induces potential unsafe problems. Therefore, it is necessary to develop a control method that effectively reduces overshoot in PMSMs. Recognizing the nature of overshoot effects, a data-driven approach, Gaussian process regression (GPR), is employed to generate the prediction. With a focus on maintaining the advantage of the GPR method, while preserving the physical properties of PMSM, an overshoot reduction-inspired motor physics embedded Gaussian Process Regression method (OR-MPE-GPR) is proposed. Inspired by the shape of the overshoot, the squared exponential (SQE) kernel function is chosen for GPR. Furthermore, by using sufficient conditions to achieve stability, the dynamic stable range and static stable range of updating rate are derived to guarantee the stability of the proposed machine learning control algorithm. Finally, comprehensive simulations and experiments compared with the state-of-the-art methods are conducted, showcasing the superior performance of the proposed method in reducing overshoot while preserving static performance within a stable region.

Recently Network Fault Prediction (NFP) has arisen new scientific interest. The ability to predict network equipment failure is increasingly identified as an effective tool to improve network reliability. This predictive capability can then be used, to mitigate or to enact preventive maintenance on incoming network failures. This could enable the emergence of zero-failure networks and allow safety-critical applications to run over larger and higher complexity heterogeneous networks. In this paper, we present a comprehensive survey on Network Fault Prediction using Machine Learning (ML) methods entirely dedicated to telecommunication networks. Specifically, we first introduce the key concepts of NFP. Second, we present the specific constraints of a Network and System Integrator (NSI), such as heterogeneity, Quality of Service (QoS), maintenance costs, and guaranteed Time To Restore (TTR) according to equipment importance, of the networks. As a consequence, we propose a new metric of ML performance, measuring the expected reduction in maintenance cost based on a ML predictive model. Third, we provide a survey of the ML methods used in NFP studies, and procure recommendations regarding their use in telecommunication networks. Fourth, we propose a summary of the different types of NFP predictions regarding different NFP problems, such as Equipment Health, Network Health, Link Failure, or Alarm Prediction. Finally, we identify the need for zero-loss and zero-delay handover, and the necessity of additional prediction elements as future research directions. In conclusion, we summarize the state of the art, and the new findings of this survey.

Partial discharges (PDs) on liquid impregnated pressboard insulation have been investigated for sinusoidal and switched voltages for two different liquids-one mineral oil and one synthetic ester. The insulation models used were a rounded electrode resting on a pressboard plate on a plane electrode representing semi-uniform field, and a sharp-edged electrode representing divergent field. The main result shows that the PD inception voltage (PDIV) is considerably lower at switched than at sinusoidal voltage. Similarly, the mineral oil showed a lower inception voltage than the synthetic ester. Finally, a lower PDIV for the sharp conical electrode compared to the spherical electrode was seen. The differences can be explained by formation and transport of space charges. The space charges will shield the high-field regions when having the same polarity as the voltage and then enhance the electric field right after a voltage polarity reversal. There is minimal time available for charge drift at polarity reversal of switched voltage. Contrary, for sinusoidal slowly varying voltage there is plenty of time for the charges to drift before the opposite voltage peak occurs. Thus, less field enhancement from space charges under sinusoidal voltage. The field enhancement due to space charges of opposite polarity is enhanced for liquids with higher injection current, such as mineral oil.

The fast and accurate interfacing of digital control platforms with power hardware constitutes a bottleneck in power electronic systems. The guidelines for the effective implementation of this interface are proprietary and not readily available to the designer. This paper presents a systematic approach for developing a mixed-domain control platform for high-frequency power converter applications. The proposed design methodology comprises three crucial steps. In the first step, the platform requirements are identified considering a wide range of switching frequency variations (up to tens of megahertz) found in power electronic applications. In the second step, a system architecture is derived, and its key building blocks and associated features are established based on the application requirements. These building blocks include serial ADCs and DACs, parallel ADCs and DACs for high-speed operation, protection circuits, PWM channels for converter gating, and digital communication features. In the third step, hardware components are selected for realizing different blocks based on a detailed comparison of the salient features of commercially available parts. A prototype of the proposed control platform is fabricated, and the experimental results are presented for validation. The results reinforce the criteria laid down in the design methodology. Further, the proposed control platform is compared with commercially available rapid control prototyping products and is shown to be economically effective for the desired features in typical power converter applications.

In this paper, we present a low-cost 5G-SA system based on edge-computing hardware with open-source components and off-the-shelf hardware. Our system is suitable for nomadic and ad-hoc 5G-SA use cases that require low power consumption and small space. We show a complete development covering how to prepare the operating system for the installation of the 5G next generation node B (gNB) and how to install and deploy the 5G-SA software components. We evaluate the performance and power consumption of our system with measurements. We show that our system achieves acceptable data rates, is very space and energy efficient with an average power consumption of 24.81 W. We provide detailed and maintained documentation on the deployment of our system and discuss possible extensions for future work.

Marine radar is widely employed in ocean monitoring systems. However, the presence of sea clutter and noise significantly hampers the detection performance of marine radar. In this paper, we propose an unsupervised-to-supervised sea clutter suppression and denoise framework. The unsupervised stage provides sufficiently high-quality pseudo-labels to the supervised stage, effectively addressing the challenge of obtaining clean labels in current supervised deep learning-based (DL-based) sea clutter suppression methods. The experiments demonstrate that our proposed method exhibits excellent adaptability in both single-target and multi-target scenarios, as well as under various pulse quantities, effectively suppressing sea clutter and noise in low signal-to-clutter ratio (SCR) and signal-to-noise ratio (SNR). In addition, we also verified that the proposed method does not lose clean signals at high SCR and SNR. To validate the widespread applicability of our unsupervised stage, we integrate it into other supervised DL-based clutter suppression methods, and the experiments show a noticeable improvement in the performance of these methods. To further verify the effectiveness of the proposed framework, we conduct validation on downstream applications of radar target detection. Our code will be released after possible acceptance.

This paper has been accepted by the "Unmanned Systems" journal. Paper link: https://doi.org/10.1142/S2301385025500839

In a world where traditional authentication systems are constrained, the introduction of voice-based authentication provides a viable option. This cutting-edge biometric security method uses unique vocal traits including pitch, tone, and speech patterns to create a unique voiceprint. Its sophisticated and dependable verification procedure, when combined with voice-activated services and secure access systems, not only solves au-thentication issues but also improves user experience overall. Our suggested method provides a thorough blueprint for developing a lightweight voice authentication system that is based on text. This system leverages an effective encoder model to convert voice spectrograms, making deployment easy even on edge devices with limited resources. Various dimensionality reduction techniques are explored to obtain optimal voice embeddings that capture speaker uniqueness while minimizing model complexity. A key novelty is the application of model compression techniques including lightweight architectures and Siamese Networks to obtain highly condensed voice embeddings for each user, reducing storage and infrastructure costs compared to traditional voice authentication methods. The proposed lightweight spectrogram embeddings leverage language-agnostic acoustic features, enabling language-independent speaker verification. Additionally, dimensionality reduction applied during voice registration allows the capturing of discriminative voice characteristics in a low-dimensional compact feature space. This significantly cuts down the storage requirements and infrastructure costs per user compared to standard voice biometric approaches, while retaining competitive verification performance. The architecture is optimized for responsiveness by leveraging lightweight frameworks. The proposed system delivers competitive voice authentication capabilities while minimizing memory, computational, and energy footprints. This makes the system useful for integration into smart devices and paves the way for ubiquitous voice biometrics.

This paper explores the problems caused by spam messages on the internet. It talks about different types of spam, like email spam and fake advertisements on social media. The paper also looks at how spam can make things difficult for people, like draining resources and tricking them into clicking on harmful links. The paper suggests fighting against spam by enabling filters in emails, using third-party spam filtering, avoiding clicking suspicious links, reporting, and blocking and avoiding using personal or professional emails. Lastly, the paper discusses what might happen with spam in the future and the challenges we may face. Overall, the paper aims to help and understand then find ways to deal with it for a safer and smoother experience.

Nearly ideal vertical AlxGa1-xN (\(0.7\ \le x<1.0\)) p-n diodes are fabricated on an AlN substrate. Distributed polarization doping (DPD) was employed for both p-type and n-type layers of the p-n junction, instead of conventional impurity doping, to overcome the major bottleneck of AlN-based material: the control of conductivity. Capacitance-voltage measurements revealed that the net charge concentration agreed well with the DPD charge concentration expected from the device layer structure. The fabricated devices exhibited a low turn-on voltage of 6.5 V, a low differential specific ON-resistance of 3 mΩ cm2, electroluminescence (maximum at 5.1 eV), and an ideality factor of 2 for a wide range of temperatures (room temperature-573 K). Moreover, the breakdown electric field was 7.3 MV cm-1, which was almost twice as high as the reported critical electric field of GaN at the same doping concentration. These results clearly demonstrate the usefulness of DPD in the fabrication of high-performance AlN-based power devices.

Emotions are considered to convey much meaning in communication. Hence, artificial methods for emotion categorization are being developed to meet the increasing demand to introduce intelligent systems, such as robots, into shared workspaces. Deep learning algorithms have demonstrated limited competency in categorizing images from posed datasets with the main features of the face being visible. However, the use of sunglasses and face masks are common in our daily lives, especially with the outbreak of communicable diseases such as the recent coronavirus. Anecdotally, partial coverings of the face reduce the effectiveness of human communication, so would this have hampering effects on computer vision, and if so, would different emotion categories be affected equally? Here, we analyze the performance of emotion classification systems when faces are partially covered with simulated sunglasses and face masks. Deep neural networks consider all pixels in an image as equally important unlike the neuroscientific findings on how humans recognize emotions. Hence, we propose a method that considers different constituent parts (e.g. mouth, eyes, and jaw) separately, giving more attention to relevant (uncovered) regions of the face. The method is compared with three standard, partial coverings-based and attention-based methods. We found that face coverings worsen emotion categorization by up to 74% for the state-of-the-art methods, whereby emotion categories are affected differently by different coverings, e.g. clear mouth coverings have little effect on categorizing happiness, but sadness is affected badly. The proposed method (on average 60.43%) has significantly improved the performance over the standard deep learning (< 46% on average), partial coverings-based (< 47% on average), as well as the attention-based (< 51% on average) methods for both the CK+, KDEF, and RAF-DB datasets when faces were partially covered with sunglasses or different face masks.

Rationale: Whether working together to move a table or supporting a child learning to ride a bike, physically connected individuals use the exchange of haptic information to improve motor performance. However, this improvement occurs at the cost of additional effort for the more skilled partner. Objective: Here, we aim to assess whether an asym-metric connection, consisting of a stiffer link to the less skilled partner, could increase performance without additional effort in collaborative tasks. Methods: Through computational modelling, we first evaluated such a hypothesis on simulated human dyads tracking a common target. The approach was then experimentally validated on a three degree-of-freedom tracking task using two commercial robots as individual interfaces. Results: The simulation and experimental results confirm that using an asymmetric connection stiffness can improve joint performance without requiring additional effort from either partner compared to solo performance. Conclusion: This suggests that the training of motor skills with a proficient partner-like a physical therapist assisting a patient or a violin teacher demonstrating bowing techniques-may be enhanced through the use of robot-mediated asymmetric haptic communication. Index Terms-Human-robot interaction, joint action, human motor control Impact Statement-Using an asymmetric connection in a dyad enables the partners to improve their common performance without additional effort.