Search

This study investigates subjects’ behavior in visual audio conflict conditions in stressful and stressless environments. To increase subjects’ stress and anxiety in subjects we placed subjects in a chaotic city and to decrease subjects’ stress we benefit from a green virtual environment. On the other hand to study the effect of sound on physiological data we designed confilict condition in which in the City environment participants heared the sound of Forest and in the Forest nvironment they heared the sound of City. To evaluate our findings we benefit from the heart rate of the participants which is directly related to subjects’ emotions and participants can not involve their opinion to change this feature autonomously. French government funding managed by the National Research Agency under the Investments for the Future program (PIA) grant ANR-21-ESRE-0030 (CONTINUUM).

Doubly-fed induction generators (DFIG) have been widely used in wind turbines installed in the last decades. These generators are prone to some faults that could deteriorate their performance and even lead to their outage from the network. Stator inter-turn short-circuits (SITSC) and high resistance connections (HRC) in the stator are two major types of faults that cause electrical asymmetry in the stator circuit. Yet, SITSC are more noticeable and require immediate scrutiny. Hence, if an HRC can be distinguished from a SITSC fault, the immediate outage of the WT can be avoided in the case of an HRC. In this paper, both types of faults are studied and compared, being their detection performed using appropriate fault indices obtained from the stator current, rotor current, and rotor modulating voltage signals, all available in the control system of the DFIG. Several fault severity indices are proposed for a better evaluation of the fault extension, and the discrimination between SITSC and HRC is discussed. The performance of the defined fault indices is verified using a magnetic equivalent circuit model of the DFIG and an experimental setup with the DFIG running at several operating conditions.

We present a Machine Learning (ML) use-case for reducing latency in wireless networks, by replacing some of the traditional and time-consuming procedures. The proposed method reduces the time needed for collecting a large number of measurements for the target cell selection during the inter-frequency handover procedure. Our paper contributes to the current literature, firstly, by showcasing how ML can replace traditional procedures in order to reduce the latency. Secondly, by presenting results based on the real measurements from a live network. Finally, by training a huge number of models over a large geographical area, where multiple base stations are deployed, reporting the performance over all those models. This contribution extends previous work in the area, that was based on simulations and improves confidence in the applicability of the proposed solution in real networks. The proposed method not only reduces the total handover delay and service disruption but can also be considered as an enabler for new verticals and use-cases, as we discuss in this paper.

Training of Artificial Neural Networks (ANNs) have been improved over the years using meta heuristic algorithms that introduce randomness into the training method but they might be prone to falling into a local minima in a high-dimensional space and have low convergence rate with the iterative process. To cater for the inefficiencies of training such an ANN, a novel neural network is presented in this paper using the bio-inspired algorithm of the movement and mating of the mayflies. The proposed Mayfly algorithm is explored as a means to update weights and biases of the neural network. As compared to previous meta heuristic algorithms, the proposed approach finds the global minima cost at far less number of iterations and with higher accuracy. The network proposed, which is named Mayfly Algorithm based Neural Network (MFANN) consists of an input layer, a single hidden layer of 10 neurons and an output layer. Two University of California Irvine (UCI) database sample datasets have been used as benchmark for this study, namely the Banknote Authentication (BA) and the Cryotherapy, for which the training accuracy achieved is 99.2350% and 96.6102%, whereas the Testing accuracy is 99.1247% and 90% respectively. Comparative study with grey wolf optimization neural network (GWONN) and particle swarm optimization neural network (PSONN) reveal that the proposed MFANN achieves 1–2% better accuracy with training dataset and 2% better accuracy with testing dataset.

Machine learning (ML) is an important tool for enabling automation in radio access networks. Multiple ML-based applications utilize the device radio-fingerprint, for example to improve network functions such as paging and inter-frequency mobility. The measurements on the plurality of beams in NR can enable the network to get an improved radio-fingerprint in comparison to previous technologies (e.g. LTE), but with the cost of increased overhead in device measurements and signaling. In this paper, the device radio-fingerprint is denoted radiolocation. The increasing use of ML-based applications with detailed radiolocation not only increases the device signaling but also increases the input dimensionality and complexity of the ML model. Hence, it is essential to propose method(s) for reducing the dimensionality by compressing the radiolocation and reducing the overall signaling by downloading such compression method(s) to the device. In this paper, one such solution is proposed, and the challenges in downloading a compression model to the device is further highlighted. Moreover, the performances for two ML applications, namely secondary carrier prediction and beam prediction, with and without compressed radiolocation are compared. The results indicate that a compressed version of radiolocation can achieve prediction performance close to the uncompressed one.

Localization has been a notable feature in wireless communications due to the increasing demand for location information. Fingerprinting-based (FP) localization methods are promising for rich scattering environments due to their high reliability and accuracy. The Gaussian process regression (GPR) method could potentially be used as an FP-based localization method to facilitate localization and provide high accuracy. However, it is limited by high complexity, especially in a large-scale environment. In this paper, we propose an FP-based localization method in collocated massive multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) systems using the affinity propagation clustering (APC) algorithm and Gaussian process regression (GPR) to estimate the user’s location. Fingerprints are extracted based on instantaneous channel state information (CSI) by taking full advantage of the high resolution in the angle and delay domains. Then, the training fingerprints are clustered using the (APC) algorithm to reduce matching complexity and computational complexity. Finally, the data distribution within each cluster is accurately modeled using GPR to provide excellent support for further localization. Simulation studies reveal that the proposed method improves localization performance significantly by reducing the location estimation error. Additionally, it reduces the matching complexity and computational complexity.

Location Based Social Networks (LBSNs) enable their user to share their check-ins and post reviews about them. The availability of spatial and textual information in LBSNs offers an opportunity to explore user’s history and preferences to find the locations that the user might be interested in. Point-Of-Interests (POIs) spatial features are one of the most important data available on LBSNs as it has a huge impact on user's choice of new location to visit. Users’ reviews and POIs’ categories are another valuable resources of information in LBSNs which help infer users’ interest and POIs’ features. Recent researches attempt to improve the performance of POI recommendation models by making use of different information sources available in social network. In this paper, we examine the impact of using this information on the accuracy of recommendation task. Our major contribution is proposing the model which use heterogeneous context information in the form of a weighted linear combination. We argue that the weights of this combination should be learned for each user separately instead of using the same set of weights for all users. We provide an algorithm for learning the weights for each user such that recommendation accuracy is improved. In addition, it is enable to incorporate extra information source to our proposed model without requirement of changing the model completely or adding extra complexity to it. Experiments conducted on two large datasets of real world, Yelp and Foursquare, shows the effectiveness of the proposed method.

Fingerprinting (FP) localization methods are used in massive multiple-input multiple-output (MIMO) systems due to their high reliability and accuracy. The Gaussian process regression (GPR) method could potentially be used, as an FP-based localization method, in a massive MIMO system to provide high accuracy. However, it is limited by high complexity, especially in a large-scale environment. In this paper, we propose an FP-based localization method, using affinity propagation (AP) clustering and Gaussian process regression (GPR) to estimate user’s location in a distributed massive MIMO system based on the uplink received signal strength (RSS) vectors. First, the training RSS vectors are clustered using the AP algorithm to reduce the computational complexity. Then, the data distribution within each cluster is accurately modeled using GPR to provide excellent support for further positioning. Simulation studies reveal that the proposed method improves root-mean-squared estimation error (RMSE) performance significantly by reducing the location estimation error compared to using only GPR for all training RSS data. Also, it reduces the computational complexity of using GPR.

We propose a low-latency approach for generating secure electrocardiogram (ECG) feature-based cryptographic keys. This is done by taking advantage of the uniqueness and randomness properties of ECG’s main features. This approach achieves a low-latency since the key generation relies on four reference-free ECG’s main features that can be acquired in short time. We call the approach several ECG features (SEF)-based cryptographic key generation. SEF consists of: 1) detecting the arrival time of ECG’s fiducial points using Daubechies wavelet transform to compute ECG’s main features accordingly; 2) using a dynamic technique to specify the optimum number of bits that can be extracted from each main ECG feature, comprising of PR, RR, PP, QT, and ST intervals; 3) generating cryptographic keys by exploiting the above-mentioned ECG features; and 4) consolidating and strengthening the SEF approach with cryptographically secure pseudo-random number generators. Fibonacci linear feedback shift register and advanced encryption standard algorithms are implemented as the pseudo-random number generator to enhance the security level of the generated cryptographic keys. Our approach is applied to 239 subjects’ ECG signals comprising of normal sinus rhythm, arrhythmia, atrial fibrillation, and myocardial infraction. The security analyses of the proposed approach are carried out in terms of distinctiveness, test of randomness, temporal variance, and using National Institute of Standards and Technology benchmark. The analyses reveal that the normal ECG rhythms have slightly better randomness compared with the abnormal ones. The analyses also show that the strengthened SEF key generation approach provides a higher security level in comparison to existing approaches that rely only on singleton ECG features. For the normal ECG rhythms, the SEF approach has in average the entropy of about 0.98 while cryptographic keys which are generated utilizing the strengthened SEF approach offer the entropy of ~1. The execution time required to generate the cryptographic keys on different processors is also examined. The results reveal that our SEF approach is in average 1.8 times faster than existing key generation approaches which only utilize the inter pulse interval feature of ECG.

In the present paper the effects of 940 MHz electromagnetic field (EMF) with two types of polarization (linear vs. circular electric field polarization) on the structure of Adult Hemoglobin (HbA) protein macromolecule are studied. The waves produced by mobile phone bases stations in urban areas are almost circularly-polarized, whereas most conventional studies have focused on linearly-polarized waves. Structural studies of irradiated HbA through intrinsic fluorescence showed that the conformation of protein was changed under both polarization types, and this change is significantly greater in circular polarization.

State-of-the-art classifiers have been shown to be largely vulnerable to adversarial perturbations. One of the most effective strategies to improve robustness is adversarial training. In this paper, we investigate the effect of adversarial training on the geometry of the classification landscape and decision boundaries. We show in particular that adversarial training leads to a significant decrease in the curvature of the loss surface with respect to inputs, leading to a drastically more "linear" behaviour of the network. Using a locally quadratic approximation, we provide theoretical evidence on the existence of a strong relation between large robustness and small curvature. To further show the importance of reduced curvature for improving the robustness, we propose a new regularizer that directly minimizes curvature of the loss surface, and leads to adversarial robustness that is on par with adversarial training. Besides being a more efficient and principled alternative to adversarial training, the proposed regularizer confirms our claims on the importance of exhibiting quasi-linear behavior in the vicinity of data points in order to achieve robustness.

The Coupling Efficiency (CE) of a beam propagating through the free space to a Single Mode Fiber (SMF) by a lens with focal length f and D aperture size is defined by the ratio of the power coupled to the fiber C > to the average power A > at the aperture plane [1]. In the presence of turbulence, the average CE depends also on the root mean square (RMS) of the wavefront phase (σ ϕ ) distortions as $\overline{CE}={8 \over {\pi^{2} D^2 w^2_a}} \left \vert \int \int_s \exp \left[- {x^2 +y^2 \over w^{2}_a} \right] ds \right\vert^{2} \ \exp [-\sigma^{2}_\phi]$ , where $w_{\alpha} = {\lambda f \over \pi w_{o}}$ is the projected mode field radius of the fiber on the aperture plane, w 0 is the radius of the optical fiber mode with integral is performed over the effective area of the receiving aperture [2].

Recently, there has been great interest in the field of metabolic engineering. Engineered microorganisms such as bacteria can be exploited to synthesize value-added bioproducts. Conversely, optimal genetic manipulations to reach to operative engineered cells is very costly; hence, regarding whole-cell functionalities, a number of proficient simulation algorithms have been developed and widely applied for systems-level analyses, and as a consequence, systems metabolic engineering. However, most of the currently-used in silico algorithms are usually time-consuming. In other words, it is requisite that computer scientists gain prior knowledge of molecular biology and biochemistry to be able to develop more effective algorithms in this field. Here, in the current study, we regard the metabolic network of interest as a general graph to present algorithmic perspective of the complex metabolic systems. A cell's metabolism is modeled in the form of a metabolic network. This way, the major problems in the field of metabolic engineering can be tackled using graph theory and graph mining algorithms. For this purpose, we reconstruct the metabolic network as a general graph, redefine the metabolic engineering problems accordingly, and finally, solve one of them as a sample.

Spectral-based fault localization is a statistical technique that aims at helping software developers to find programming faults quickly. In this technique fault locations are found by analyzing abstractions of program traces to provide a ranked list of most probable faulty components (e.g., program statement). Although the semantic structure of code is not considered in Spectral-based fault localization methods, they are widely known as high-performance and state-of-the-art methods. Many studies have been conducted in this area that have led to the creation of various methods, with different accuracy and efficiency. Nevertheless, the relative performance of these methods are neither examined nor discussed in the literature. In this paper, we introduce an empirical framework for implementation of SBFL methods in large-scale real bug java programs. Implementation results are reported using evaluation metrics to provide a clear comparison of the performance of the most effective methods.

Seamless wireless connectivity and low latency com- munication in mobility scenarios are two fundamental require- ments in the fifth generation networks to effectively cover a broader range of use cases such as massive machine type communications and ultra-reliable low-latency communications for new industrial/potential users, as well as enhanced mobile broadband for conventional users. In this paper, we aim at measuring certain key performance indicators in mobility scenarios, highlighting some practical inefficiencies that require more careful consideration when de- signing and enhancing mobility features of 5G networks. We investigate the round trip time, uplink throughput and the impact of handover interruption time on these metrics. Our results are based on three months experiment campaign in a mix of rural, suburban, and urban environment using the geographically distributed MONROE platform. Based on the observed discrep- ancies between measured and standardized LTE performance, we conclude the paper with a discussion of techniques that may need more attention for mobility scenarios in fifth generation mobile networks. DigitalWell Research Fas 1, 5465

Modular converters with stacked cells have a widely-known advantage of achieving higher voltage or current levels by using standard low-voltage or low-current switches. This study presents a galvanically isolated and scalable photovoltaic inverter with stacked output cells. There is no dc link with electrolytic capacitors in this three-phase inverter. Instead, a distributed ac link composed of small ac capacitors and transformer magnetizing inductances is proposed. The transformer core is utilized symmetrically in positive and negative half cycles. Due to non-appearance of a dc link, fast dynamic response is achievable. The ac link provides soft-switching for all switches in the entire load range. The energy stored in transformer winding leakage inductances is fed to the ac capacitors to mitigate switching spikes without using auxiliary snubbing switches. Current and voltage symmetry among the cells is provided naturally. The transformer design is simple because the inverter operation is not affected by unequal winding leakage inductances. An analytical study and design curves are provided to determine the ac link components. Experimental validation verifies the operation of a 750 W prototype.

Recent studies have illustrated that social networks are valuable sources of information which can be used for various purposes. In recommender systems, researchers have been motivated to utilize social information not only to enhance the prediction accuracy, but also to address the data sparseness challenges such as cold start problem. In this paper, we propose a social memory-based Collaborative Filtering (CF) approach and examine the impact of integrating social ties on both accuracy and prediction coverage. A simple learning method is used to tune the contribution of social information, in the proposed similarity measure, as a weight parameter. We compared our experimental results with several state-of-the-art memory-based and model-based CF approaches using Epinions dataset to investigate the performance of our method in extreme sparsity of real-world data. Results show that our approach provide high accuracy as well as reasonable prediction coverage comparing with other methods.