Classification of mental tasks from electroencephalogram (EEG) signals play a crucial role in designing various brain-computer interface (BCI) applications. Most of the current techniques consider each channel as independent, neglecting the functional connectivity of the brain during mental activity and are primarily subject specific. This paper proposes a graph signal representation to classify a pair of mental tasks using multichannel EEG signals (MTMC-EEG) with cross subject classification within the database. Here, each channel of EEG signal corresponds to nodes of the task based graph whose EEG time series resides on the respective nodes. Functional connectivity of the brain between these nodes is obtained using smoothness constraint based Graph Signal Processing (GSP) technique. Graph spectral features namely, two-norm total variation of eigen vector (TNTV) corresponding to weighted adjacency matrix, graph Laplacian energy (GLE) using eigenvalues of Laplacian matrix and convex sum of TNTV and GLE in the form of joint total variation energy (JTVE) are proposed in this paper. The performance of the proposed methodology is evaluated on publicly available two different databases of MTMC EEG signals using benchmark classifiers and compared with the state of the art. Further, the superiority of the proposed metric obtained from the smoothened graph of GSP technique is validated by comparing it with Pearson correlation and Gaussian radial basis function (RBF) based functional connectivity in terms of accuracy, F-Score, and information transfer rate (ITR). The robustness of the proposed method is validated by adding white Gaussian noise (AWGN) to the EEG signals using different SNRs.

In this paper, we continue the analysis of electrooptical studies of the prebreakdown stage in water in the submicrosecond range. We consider a cathode streamer developing in a system of the spherical electrodes. Computer simulation of electric showed that at a constant e = 80, the field strength at the head of the cathode streamer should not exceed E = 1.9 MV/cm. Taking into account the nonlinearity of the permittivity e = e(E) made it possible to estimate this field strength as E ~ 2.4-3.1 MV/cm. It is shown that even with some excess of the strength over the threshold value (ES ~ 3 MV/cm), at which the state of dipole saturation is possible; its consideration does not lead to a noticeable change in the result. Thus, a significant decrease in the field strength near the cathode streamer compared to the anode streamer was confirmed.

Navigation bronchoscopy is a minimally invasive procedure in which doctors pass a bronchoscope into a subject’s airways to sample the target pulmonary lesion. A three-dimensional (3D) airway roadmap reconstructed from Computer Tomography (CT) scans is a prerequisite for this procedure, especially when the target is distally located. Therefore, an accurate and efficient airway segmentation algorithm is essential to reduce bronchoscopists’ burden of pre-procedural airway identification as well as patients’ discomfort during the prolonged procedure. However, airway segmentation remains a challenging task because of the intrinsic complex tree-like structure, imbalanced sizes of airway branches, potential domain shifts of CT scans, and few available labeled images. To address these problems, we present a deep learning-based pipeline, denoted as NaviAirway, which finds finer bronchioles through four major novel components – feature extractor modules in model architecture design, a bronchiole-sensitive loss function, a human-vision-inspired iterative training strategy, and a semi-supervised learning framework to utilize unlabeled CT images. Experimental results showed that NaviAirway outperformed existing methods, particularly in identification of higher generation bronchioles and robustness to new CT scans. On average, NaviAirway takes five minutes to segment the CT scans of one patient on a GPU-embedded computer. Moreover, we propose two new metrics to complement conventional ones for a more comprehensive and fairer evaluation of deep learning-based airway segmentation approaches. The code is publicly available on https://github.com/AntonotnaWang/NaviAirway.

In this original study, we investigate the performances of machine learning algorithms on a neonatal sepsis detection task. We consider this work to be of great interest to both engineers and clinicians, as it uses non-invasive, already existing, vital signs monitoring signals in a population of very low birth weight infants at high risk of sepsis. Vital sign variability may indeed represent a general indicator of health and wellbeing and be helpful in the early detection of systematic inflammation such as sepsis. We used state of the art feature extraction technics and evaluate a large variety of binary classification models among which a neural network based generative model. The models were chosen from two main families: discriminative and generative. This enables a comprehensive study of different kinds of traditional and advanced binary classification algorithms. Our study reveals that advanced machine learning models are more robust to changes in the feature extraction pipeline, although linear classifiers have a comparable performance when the feature extraction is tuned. The advanced model performing the best is a neural network based generative model which is a hybrid generative and discriminative model. A large window length when computing the features is beneficial to almost all algorithms, indicating the relevance of frequency domain related features for the neonatal sepsis detection task. Overall we obtain a classification AUROC above 0.85, which makes our prediction models potentially relevant in clinical practice. This will enable earlier therapeutic interventions and thereby reduce morbidity and mortality in infants.

Decentralization and high penetration of smart devices in IoE-enabled smart grids face the power system with complex scheduling problems. Engaging with big data produced by the interconnected infrastructures, besides the high dimensional and uncertain environment, make traditional methods incapable of addressing these problems since exact modeling of the environment under uncertainties is impracticable. Also, learning-based methods suffer from excessive complexity and the curse of dimensionality. This research proposes a Probabilistic Delayed Double Deep Q-Learning (P3DQL) which is a combination of the tuned version of Double Deep Q-Learning (DDQL) and Delayed Q-Learning (DQL). The planned algorithm makes a trade-off between overestimation and underestimation biases guaranteeing efficiency regarding sample complexity and learning proficiency by applying a delay in updating the rule. Finally, the proposed algorithm is tested on real-world data from Pecan Street Inc., assessing the performance of the P3DQL regarding peak clipping, decreasing Peak to Average Power Ratio (PAPR), and cost reduction. The results indicate the superiority of the developed algorithm over other utilized methods by 28.2% peak clipping, 12.9% PAPR decrease, and 29.4% cost reduction.

Lognormal Probability Density Functions (pdfs) are often empirically found to fit the distribution of failure lifetimes for electronic components, but prior to this work, a physical basis has not be found for why Lognormals should be applicable. Using long-thin metal interconnects atop non-conducting substrates as a representative system, we show Lognormals pdfs for device lifetimes naturally arise, whenever there is a standard Gaussian variation in device geometry, with the case studied being variations in metal interconnect line-widths.

Electricity prices determined from economic dispatch without considering fixed costs may cause high uplift payments. With fixed costs, however, a price is not a monotonic function of demand, affecting market transparency. To overcome these, convex hull (CH) pricing has recently been introduced for unit commitment with fixed costs. Several CH pricing methods were presented, and a feasible cost was used to quantify the CH price quality. The associated difficulties are 1. high computational effort required to obtain a feasible cost and 2. the associated duality gap may not be tight enough to provide accurate measure. In this paper, a novel measure to quantify the quality of CH prices is presented by establishing an upper bound to the optimal dual value approaching it from above. Near-optimal CH prices are efficiently obtained by using Surrogate Lagrangian Relaxation (SLR), meanwhile, the upper bound decreases fast due to convergence of SLR. Testing results on the IEEE 118-bus system without transmission capacities indicate that the novel quality measure reaches a value of less than 0.1% in seconds – much more accurate and faster than the measure provided by a feasible cost – demonstrating the high quality of the upper bound and the efficiency of SLR.

Despite the significant benefits that the Internet of Healthcare Things (IoHT) offered to the medical sector, there are concerns and risks regarding these systems which can delay their wide deployment as they handle sensitive, and often life-critical medical information. In addition to the IoHT concerns and risks, there are security constraints which include hardware, software, and network constraints that pose a security challenge to these systems. Therefore, security measures need to be deployed that can overcome the concerns, mitigate the risks, and meet the constraints of the IoHT. For these reasons, a subclasses intrusion detection system for the IoHT is proposed in this research work based on a novel variation of the standard One-class Support Vector Machine (OSVM), namely Deep Subclass Dispersion One-class Support Vector Machine (Deep SDOSVM), which considers subclasses in the target class, i.e. normal class, in order to minimize the data dispersion within and between subclasses, thereby improving the discriminative power and classification performance of the intrusion detection system. A deep clustering model is used for subclasses generation in the proposed Deep SDOSVM approach, namely the Dynamic Autoencoder Model (DynAE), to overcome the drawbacks of the classical clustering algorithms, and further enhance the classification performance of the intrusion detection system. The proposed deep clustering subclasses intrusion detection system was evaluated on the real-world TON_IoT dataset, and compared to other state-of-the-art one-class classifiers. Experimentation results showed that the proposed approach outperformed the other relevant one-class classifiers for network intrusion detection.

This paper compares different strategies used to enhance stability properties of grid-following voltage-source converters (VSCs). The focus is on high-frequency properties and performance is examined even above the Nyquist frequency. Due to digital delays, VSC admittance exhibits a non-passive zone above a certain frequency, which introduces negative damping and may destabilize the grid-connected operation. It is shown that typically used active damping (AD) strategies only bring positive impact up to a certain frequency, while deteriorating admittance properties in the high-frequency range. Multi-sampled pulsewidth modulation (MS-PWM) greatly extends the passive admittance region, using a single-stage current-control loop. Experimental admittance measurements are performed on a single-phase VSC, up to twice the switching frequency. Subsequently, different grid-connected scenarios are tested to show that MS-PWM retains stable operation where AD methods cause instability. In the end, noise propagation is measured and results confirm that, while AD does not bring high sensitivity, MS-PWM offers additional noise suppression. The results of this paper clearly show that MS-PWM is a better candidate for guaranteeing stability of grid-connected VSCs.

This paper deals with safe human-robot collaboration in the context of speed and separation monitoring paradigm. The core of the approach is to continuously track the separation distance between the robot and the human. The robot speed is then adjusted according to the perceived distance so that it will be able to stop before eventually come into contact with the human. We present an approach that aims at maximizing the productivity of the robot, i.e., its speed, while keeping the prescribed safety requirements satisfied. The method is based on explicit representation of danger zones -- regions around the robot, where safety requirements are violated. The motion is then generated such that the robot moves as fast as possible, while its danger zone still does not collide with human operators. The approach is validated within an experimental study.

A document by Mastane Achab . Click on the document to view its contents.

We examine the soundness of the idea and practice of engineering accreditation and look critically at the process as currently implemented in some regions around the world. We suggest areas for improvement and propose guidelines for a revised accreditation framework that addresses some of the challenges of the current system.

One of the main challenges in developing embedded radar-based gesture recognition systems is the requirement of energy efficiency. To facilitate this, we present an embedded gesture recognition system using a 60 GHz frequency modulated continuous wave radar using spiking neural networks (SNNs) applied directly to raw ADC data. The SNNs are sparse in time and space, and event-driven which makes them energy-efficient. In contrast to the previous state-of-the-art methods, the proposed system is only based on the raw ADC data of the target thus avoiding the overhead of performing the slow-time and fast-time Fourier transforms (FFTs). Furthermore, the pre-processing slow-time FFT is mimicked in the proposed SNN architecture, where the proposed model processing speed of 12 ms advances the state-of-the-art by a factor of 17.7. The experimental results demonstrate that despite the simplification the proposed implementation achieves recognition accuracy of 98.1 %, which is comparable to the conventional approaches.

Memristor-based crossbar architecture emerges as a promising candidate for 3-D memory and neuromorphic computing; however, the sneak current through the unselected cells becomes a fundamental roadblock for their development as it results in misreading and high power consumption. In this regard, we investigate Pt/Ti/NbO2/Nb2O5−x/Pt-based self-selective memristor, which combines inherent nonlinearity of NbO2 switching layer and non-volatile operation of Nb2O5−x memory layer in a single device. The results show that Pt/Ti/NbO2/Nb2O5−x/Pt memristor offers the sneak current of 310 nA, selectivity of around 174, and on/off current ratio of 4.79 compared to the sneak current of around 70 µA, selectivity of about 4.02, and on/off current ratio of around 1.55 for Pt/Ti/Nb2O5−x/Pt-based memristor. Our self-selective memristor minimizes the sneak current, but a small on/off current ratio limits their readout margin and power efficiency for crossbar array size greater than 212. Further, we demonstrate that breaking down a large-scale crossbar array into smaller subarrays and separating them by transistor switches is a more efficient way of eliminating the sneak current. The memristor-transistor split crossbar architecture essentially provides the readout margin and power efficiency of subarray for large size crossbar array and allows high-density array integration over 1Transistor-1Memristor architecture.

In this paper, an approach for helping undergraduate students understand complex algorithms through structured flowcharts (SFCs) and their implementation using graphical programming (GPR) on LabVIEW and myRio is presented. Space vector pulse width modulation (SVPWM) is an important algorithm for students studying power electronics. But, when studying this algorithm, a lot of students often find difficulty in comprehending it. In this paper, we have presented a method which helps students understand SVPWM easily and enables them to implement it on hardware with less difficulty. This article also intends to show the effectiveness of LabVIEW and myRio as constructive educational tools. The proposed approach can also be applied to teaching other complex algorithms. The proposed approach makes use of SFCs to help students develop a theoretical foundation of SVPWM in a theory class. The theory session is followed by the practical implementation of SVPWM using GPR on LabVIEW and testing it on a test-bed (a 3-phase motor connected to myRio) in two laboratory session.

This paper provides an in-depth review about opinions and applications in the new area of Human Centered Artificial Intelligence

In this paper, we propose a novel neural network for Compressed Sensing (CS) application: the Dictionary Learning based Compressed Sensing neural Network (DL-CSNet). It is fairly simple but highly effective, which consists of only three layers: 1) a DL layer for latent sparse features extraction; 2) a smoothing layer via Total Variation (TV) like constraint; and 3) a CS acquisition layer for neural network training. In particular, the TV-like smoothing layer is a perfect complement to the sparsity-oriented DL layer to achieve smooth images. The trained DL-CSNet can learn the optimal dictionary matrix so that images can be reconstructed in high quality. At last, extensive experiments have been carried out on binary images and compared to most classical CS algorithms, which shows the superior performance of the proposed DL-CSNet.

Monolithic pixel sensors integrate the sensor matrix and readout in the same silicon die, and therefore present several advantages over the more largely used hybrid detectors in high-energy physics. In this paper, a low-power, radiation-hard frontend circuit for monolithic pixel sensors, designed to meet the requirements of low noise and low pixel-to-pixel variability, the key features to achieve high detection efficiencies, is presented. The sensor features a small collection electrode to achieve a small capacitance (< 5 fF) and allows full CMOS in-pixel circuitry. The circuit is implemented in the 180 nm CMOS imaging technology from the TowerJazz foundry and integrated in the MALTA2 chip, which is part of a development that targets the specifications of the outer pixel layer of the ATLAS Inner Tracker upgrade at the LHC. One of the main challenges for monolithic sensors is a radiation hardness up to 1015 1 MeV neq/cm2 Non-Ionizing Energy Loss (NIEL) and 80 Mrad Total Ionizing Dose (TID) required for this application. Radiation source and charge injection tests up to 3 ° 1015 1 MeV neq/cm2 and 100 Mrad were performed on the MALTA2 sensor and front-end circuit, which still show good performance even after these levels of irradiation, promising for even more demanding applications such as the future experiments at the HL-LHC.