Search

The use of Internet of Things (IoT) is becoming increasingly common in agribusiness to increase food production capacity for the expanding global population. Recently, low-power wide-area networks (LPWANs) have been used in the development of IoT applications that require low power consumption and low data transmission rates. LoRaWAN is considered the most suitable communication network for LPWANs for IoT applications in smart agriculture. In this paper, we present an in-depth study of the performance of the LoRaWAN communication network in the context of an IoT application for a pilot farm. We consider several scenarios and analyze simulation results by using Network Simulator 3. We then propose a mathematical model that precisely predicts the successful packet delivery rate for this type of network considering the number of nodes and the transmission interval duration. Finally, we validate the results of our model by comparing them with other simulation results under different scenarios.

Recognition systems using multimodal biometrics attracts attention because they improve recognition efficiency and high-security level compared to the unimodal biometrics system. In this study, the authors present a secure multimodal biometrics recognition system based on the deep learning method that uses convolutional neural networks (CNNs). The authors propose two multimodal architectures using the finger knuckle print (FKP) and the finger vein (FV) biometrics with different levels of fusion: the features level fusion and scores level fusion. The features extraction for FKP and FV are performed using transfer learning CNN architectures: AlexNet, VGG16, and ResNet50. The key step aims to select separate features descriptors from each unimodal biometrics modality. After that, the authors combine them using the proposed fusion approaches were support vector machine or Softmax applies as classifiers to increase the proposed system security. The efficiency of the proposed algorithms is tested using publicly available biometrics databases. The experimental results show that the proposed fusion architectures achieve an accuracy of 99.89% and an equal error rate of 0.05%. The obtained results indicate that the biometrics recognition system using deep learning is secure, robust, and reliable.

International audience; Vessel structures such as retinal vasculature are important features for computer-aided diagnosis. In this paper, a probabilistic tracking method is proposed to detect blood vessels in retinal images. During the tracking process, vessel edge points are detected iteratively using local grey level statistics and vessel's continuity properties. At a given step, a statistic sampling scheme is adopted to select a number of vessel edge points candidates in a local studying area. Local vessel's sectional intensity profiles are estimated by a Gaussian shaped curve. A Bayesian method with the Maximum a posteriori (MAP) probability criterion is then used to identify local vessel's structure and find out the edge points from these candidates. Evaluation is performed on both simulated vascular and real retinal images. Different geometric shapes and noise levels are used for computer simulated images, whereas real retinal images from the REVIEW database are tested. Evaluation performance is done using the Segmentation Matching Factor (SMF) as a quality parameter. Our approach performed better when comparing it with Sun's and Chaudhuri's methods. ROC curves are also plotted, showing effective detection of retinal blood vessels (true positive rate) with less false detection (false positive rate) than Sun's method

International audience; The paper deals with the monolithic integration of a power converter phase leg. It focuses on the integration of a power phase leg consisting of a VDMOS and an IGBT. This association is suitable for monolithic power integration in silicon. Different monolithic chips integrating phase legs were proposed and discussed in this paper. Interesting advantages can be brought by the monolithic integration of the converter phase leg. These advantages include power chips realization simplification, control of gate transistors with respect to constant voltages (DC-link or ground-0V).

International audience; Deep learning models have strong abilities in learning features and they have been successfully applied in hyperspectral images (HSIs). However, the training of most deep learning models requires labeled samples and the collection of labeled samples are labor-consuming in HSI. In addition, single-level features from a single layer are usually considered, which may result in the loss of some important information. Using multiple networks to obtain multi-level features is a solution, but at the cost of longer training time and computational complexity. To solve these problems, a novel unsupervised multi-level feature extraction framework that is based on a three dimensional convolutional autoencoder (3D-CAE) is proposed in this paper. The designed 3D-CAE is stacked by fully 3D convolutional layers and 3D deconvolutional layers, which allows for the spectral-spatial information of targets to be mined simultaneously. Besides, the 3D-CAE can be trained in an unsupervised way without involving labeled samples. Moreover, the multi-level features are directly obtained from the encoded layers with different scales and resolutions, which is more efficient than using multiple networks to get them. The effectiveness of the proposed multi-level features is verified on two hyperspectral data sets. The results demonstrate that the proposed method has great promise in unsupervised feature learning and can help us to further improve the hyperspectral classification when compared with single-level features.

Deep learning target detection based on faster regions with convolutional neural network (Faster R-CNN) features has been applied in image processing successfully, however, it is rarely introduced to the field of hyperspectral image (HSI) target detection due to the tensor characteristics and the lack of training samples of HSI data. In this paper, the target detection based on Faster R-CNN is proposed to HSI with data set adjustment and parameter turning. As a typical tensor data, HSIs contain two-dimensional (2-D) spatial information and one dimensional (1-D) spectral information. It contains more information than ordinary images, and has unique advantages in the field of ground object and sea target detection. Therefore, the original HSI is firstly adjusted to the data set format required by the model, and the final Faster R-CNN sample data set can be achieved by combining the data set of Google Earth images. Next, a Faster R-CNN network suitable for HSI data could be built. Finally, to improve the accuracy of target detection, some parameters of Faster R-CNN would be tuned. The numerical results show that the method has the potential advantages of high precision and high speed in HSI target detection, and will have broad application prospects.

International audience; The purpose of this work is to perform co-engineering of an object classification system including both a radar sensor and a software of image processing. We aim at the smallest possible false recognition rate, considering three classes of imaged objects. For this we retain five relevant parameters which impact the recognition performances. We adopt the mixed grey wolf optimizer to provide the best set of parameters.

To deal with the complex design issues of Dynamically Reconfigurable Systems-on-Chip (DRSoCs), it is extremely relevant to raise the abstraction level in which models are expressed. A high abstraction level allows great flexibility and reusability while bypassing low-level implementation details. In this context, model-driven engineering (MDE) provides support to build and transform precise and structured models for a particular purpose at different levels of abstraction. Indeed, high-level models are successively refined to low-level models until reaching the executable ones. Thus, this paper presents an MDE-based framework for DRSoCs design enabling the transformation of UML/MARTE specifications to SystemC/TLM implementation. To achieve a high degree of expressiveness for modeling dynamic reconfiguration, we use a suitable software engineering approach based on service-oriented component architecture. Since MARTE does not cover the common features of dynamic reconfiguration domain and service orientation concepts, new stereotypes are created by refinement to add missing capabilities to the profile. Likewise, SystemC does not provide native support for dynamic reconfiguration, thus leading us to adopt a design pattern based solution for DRSoCs implementation in compliance with standards. The proposed framework is validated through a reconfigurable active 3-way crossover case study in which we demonstrate the practicability of the approach by gradual model transformations with reduced implementation effort and significant design productivity gain.

Integration of FPGAs in datacenters might have different motivations from acceleration to energy efficiency, but the goal of better performance tops all. FPGAs are being utilized in a variety of ways today, tightly coupled with heterogenous computing resources, and as a standalone network of homogenous resources. Open source software stacks, propriety tool chain, and programming languages with advanced methodologies are hitting hard on the programmability wall of the FPGAs. The deployment of FPGAs in datacenters will neither be sustainable nor economical, without realizing the multi-tenancy in multiple FPGAs. Inter-FPGA communication among multiple FPGAs remained relatively less addressed problem this area. In this proposed study, we made a theoretical case as a potential solution for this bottleneck problem to realize multi-tenant multi-FPGAs in datacenters. The proposed scheme poses a challenge as well as an opportunity for the scientific community to contribute.

The crowd counting task involves the issue of security, so now more and more people are concerned about it. At present, the most difficult problem of population counting consists in: how to make the model distinguish human head features more finely in the densely populated area, such as head overlap and how to find a small-scale local head feature in an image with a wide range of population density. Facing these challenges, we propose a network for multiple feature convolutional neural network, which is called MFNet. It aims to get high-quality density maps in the high-density crowd scene, and at the same time to perform the task of the count and estimation of the crowd. In terms of crowd counting, we use multiple sources of information, that is HOG, LBP and CANNY. With the support vector machine (SVM), each source provides us not merely a separate count estimation, but other statistical measures. In order to effectively solve the problem of extracting scale-related features in crowd counting, we have integrated MFNet, a convolutional neural network architecture. By comparing the experimental results of multiple data sets, MFNet is superior to other population counting methods.

Tensors are high order muti-way arrays which extend vectors and matrices to higher, greater than three, dimensions. Rather than vectorizing data, tensors allow its representation in its original raw form, hence offering better understanding and interpretation of the underlying data structure. In this paper, we propose an activity recognition and fall detection approach which makes use of tensors as the mean for wearable inertial sensors data representation. To reduce the tensor size and only keep important information, we project the tonsorial data into a lower dimensional space using multilinear principle component analysis. To evaluate the proposed approach, we carry out experiments on a publicly available large fall and activities dataset. The results show the defectiveness of tensorizing data compared to its processing through concatenated vectors.

International audience; In this work, we propose a new concept of fail-safe switching cells based on the use of integrated fuses. These fuses allow the circuit to be partially isolated and reconfigured, which therefore becomes fault tolerant and prevents total shut down. We focused our work on the design and 3D simulation by finite elements method of monolithic fuses integrated on silicon substrate, taking into account static and dynamic specifications. Thermal management in steady state is improved by dielectric epoxy thermal insulation under each constriction of the fuse. Implementation and preliminary practical tests at medium voltage are reported. The effect of a coating using silicone gel around the fuses is analyzed to improve the cut-off capability.

International audience; Modern datacenters are reinforcing the computational power and energy efficiency by assimilating field programmable gate arrays (FPGAs). The sustainability of this large-scale integration depends on enabling multi-tenant FPGAs. This requisite amplifies the importance of communication architecture and virtualization method with the required features in order to meet the high-end objective. Consequently, in the last decade, academia and industry proposed several virtualization techniques and hardware architectures for addressing resource management, scheduling, adoptability, segregation, scalability, performance-overhead, availability, programmability, time-to-market, security, and mainly, multitenancy. This paper provides an extensive survey covering three important aspects—discussion on non-standard terms used in existing literature, network-on-chip evaluation choices as a mean to explore the communication architecture, and virtualization methods under latest classification. The purpose is to emphasize the importance of choosing appropriate communication architecture, virtualization technique and standard language to evolve the multi-tenant FPGAs in datacenters. None of the previous surveys encapsulated these aspects in one writing. Open problems are indicated for scientific community as well.

Classification is a key technique in hyperspectral image (HSI) applications. Deep learning algorithms, which exhibit strong modeling and representational capabilities, have been successfully adopted in fields such as image and language processing. And convolutional neural networks (CNNs) have been used for HSI classification and some interesting results have been obtained. Owing to local connection and weight sharing, the number of parameters is reduced to some extent, but there are still many parameters and the deeper the network, the larger is the number of parameters. The network performance is strongly influenced by the parameter settings. To obtain the optimal CNN parameters for HSI classification, this paper proposes a classification method based on a CNN with parameter tuning (CNN-PT). The network parameters are tuned in turn according to the unique variable principle. Simulation results show that the proposed CNN-PT method has considerable potential for HSI classification compared to previous methods.

Accurate classification is one of the most important prerequisites for hyperspectral applications and feature extraction is the key step of classification. Recently, deep learning models have been successfully used to extract the spectral-spatial features in hyperspectral images (HSIs). However, most deep learning-based classification methods are supervised and require sufficient samples to guarantee their performance. And the labeled samples in HSI are limited. To solve this problem, unsupervised feature extraction based on improved Wasserstein generative adversarial network (WGAN) is proposed in this paper. Further, in order to fully explore the spectral-spatial features in HSIs, a three-dimensional (3D) model is designed. Considering that its difficult for GANs to generate high dimension data, the dimension of HSI is reduced firstly by principle component analysis (PCA). Then, an improved WGAN is trained unsupervised with HSIs after dimensionality reduction to achieve the stable status. Finally, the discriminator of the trained improved WGAN is used as a feature extractor and followed by a classifier, which can be used to classify the HSIs. In order to evaluate the performance of the proposed method, PCA-based methods and GAN-based methods are compared in the experiment of a real-world HSI. Experimental results have shown that our proposed method gained more promising results which has great potential prospects in HSI classification.

Electrical impedance tomography is a non-invasive imaging technique capable of void fraction measurements of two-phase flows at high-acquisition rates. This paper presents the design features of an in-house developed prototype fast EIT system and a 16 electrodes sensor. With simplified system architecture, the voltage of the excitation signal is controlled and the current is measured. The proposed full scan data collection strategy opens up new prospects for novel flow regime identification and image reconstruction techniques currently under development. A new calibration procedure that is based on the data from full scan strategy provides reliable absolute measurements in applications, where reference measurements are not available or reliable (e.g., gas-liquid flows with drifts of liquid phase temperature). Experimental results are compared with the analytical and numerical solutions. Fundamental studies for understanding the contact impedance give insight on the influence of important parameters: excitation frequency and liquid conductivity. The quality of the measurements at different frame rates is analyzed. The repeatability at 80 kHz of the prototype highlights similar figure of merit than reported fast EIT systems. The increase in frame acquisition rate allows up to 6250 frames per second for adjacent strategy and up to 833 frames per seconds for full scan strategy.

International audience; Design of Systems-on-Chip has become very common especially with the remarkable advances in the field of high-level system modeling. In recent years, Matlab also offers a Simulink interface for the design of hardware systems. From a high-level specification, Matlab provides self-generation of HDL codes and/or FPGA configuration codes while providing other benefits of easy simulation. In addition, a large part of the Systems-on-Chip use at least one image processing algorithm and at the same time border detection is one of the most used algorithms. This paper presents a study and a hardware implementation of various algorithms of borders detection realized under Xilinx System-Generator. The various algorithms are implemented using Xilinx FPGA device, the simulation and synthesis results are also compared. We use the Xilinx Zed-Board for physical implementations in-the-loop.

International audience; Power electronic converters (power modules) are essential devices for high efficient power management. Applications are ever more demanding in terms of compactness, reliability and cost reduction with increased intrinsic performance. The current technology of standard power modules, that use IGBT chips, is reaching its limits and the margins for improvement do not allow to meet the short-term emerging needs. Indeed, in a standardpower module, each switching cell requires the manufacture of two chips and two wirebonds for interconnections. Each wirebond is a source of stray inductance and constitutes a limiting factor of electrical performance, a source of an increase in the level of electrical stress and of reliability problems. It is customary to place a decoupling capacitor on the power supply terminals of the switching cell. This capacitor makes it possible to compensate the stray inductance prior to the switching cell but itdoesn’t compensate in any way the stray inductance within the switching cell. This paper proposes for the first time a concept of an ultimate monolithic wire-bondless switching cell that allows to suppress the wirebonds within the switching cell and therefore minimize the stray inductance value. Consequently, it is suited for the emerging applications requiring to achieve simultaneously: reliability, compactness, intrinsic performance and reduced manufacturing cost.

The investigation of quickly-evolving flow patterns in high-pressure and high-temperature flow rigs requires the use of a high-speed and non-intrusive imaging technique. Electrical Impedance Tomography (EIT) allows reconstructing the admittivity distribution characterizing a flow from the knowledge of currents and voltages on its periphery. The need for images at high frame rates leads to the strategy of simultaneous multi-frequency voltage excitations and simultaneous current measurements, which are discriminated using fast Fourier transforms. The present study introduces the theory for a 16-electrode simultaneous EIT system, which is then built based on a field programmable gate array data acquisition system. An analysis of the propagation of uncertainties through the measurement process is investigated, and experimental results with fifteen simultaneous signals are presented. It is shown that the signals are successfully retrieved experimentally at a rate of 1953 frames per second. The associated signal-to-noise ratio varies from 59.6&ndash, 69.1 dB, depending on the generated frequency. These preliminary results confirm the relevance and the feasibility of simultaneous multi-frequency excitations and measurements in EIT as a means to significantly increase the imaging rate.

Carbon monoxide (CO) is a particularly toxic gas, and the CO poisoning is the leading cause of fatal poisoning in the world. One solution to deal with this dangerous gas is to exploit the advantages given to us by the applications and the technology advances in internet of things (IoT). In this context, we propose to integrate carbon monoxide detection objects to the internet, as part of civil protection intervention applications in case of detection of carbon monoxide to avoid poisoning. To do this, we propose a typical IoT architecture to manage the alerts sent to civil protection in the event of detection of carbon monoxide (CO). The architecture is based on the Message Queue Telemetry Transport (MQTT) communication protocol. Finally, we present the realization of our Test-bed and its components to integrate the object of detection of carbon monoxide on the internet.

One of the objectives of image processing is to detect the region of interest (ROI) in the given application, and then perform characterization and classification of these regions. In HyperSpectral Images (HSI) the detection of targets in an image is of great interest for several applications. However, edge detection is not considered by the community. Moreover, it is less easy to define what can be a contour in a HSI than in a conventional image where the edges are defined as a rapid change in the intensity of the pixels. In fact, because the spectral data (hyperspectral pixels) consist of several hundreds of values, it becomes more difficult to define the variations between two pixels. One can found in the literature a few edge detection works on multi or hyperspectral images, based on statistical similarity criteria (maximum likelihood, Mahalanobis distance), or based on a geometric measurement approach (study of distances or angles between pixel vectors), some works also adapt to multispectral data some classical convolution filters (Sobel gradient, etc.), or mathematical morphology methods. But all these methods are more suited to the multispectral than the hyperspectral data because they are put in difficulty by the large amount of information known as Hugues phenomenon. In this paper we propose a new multidimensional method, based on tensor modelling of HSI, and using a local approach thanks to subtensors rank estimation to perform edge detection. This could be a good way to select regions of interest which are known to be useful when small target detection is needed. Generally, when ROI containing targets is previously selected, the detection results are better. The ROI selection for large HSI containing objects with large and small sizes, is based on tensorial modelling, and an estimation of local signal subspace rank variations. Examples of the results obtained by the proposed method show that the local variations of signal subspace rank for subtensors of 5x5 pixels on 143 bands centered on each pixel of the image. The processed real-world HSI containing an almost uniform background and 4 rows of different sizes of panels. The obtained results concerning the evolution of the local rank of subtensors centered on each pixel of the HSI, show that the local rank of the signal subspace and the panels are linked. According to these promising and convincing results, in this paper we propose an algorithm for edge detection based on local rank variations.

International audience

This paper presents a new approach for face and kinship verification under unconstrained environments. The proposed approach is based on high order tensor representation of face images. The face tensor is built based on local descriptors extracted at multiscales. Besides, we formulate a novel Multilinear Side-Information based Discriminant Analysis (MSIDA) to handle the weakly supervised multilinear subspace projection and classification. Using only the weak label information, MSIDA projects the input face tensor in a new subspace in which the discrimination is improved and the dimension of each tensor mode is reduced simultaneously. Experimental evaluation on four challenging face databases (LFW, Cornell KinFace, UB KinFace and TSKinface) demonstrates that the proposed approach significantly outperforms the current state of the art.

With the rapid advent of using various devices like smart phones, vehicles etc, the connection of these devices with the help of Internet connectivity has emerged to IoT paradigm. The interconnection of smart objects under the various real world constraints like communication technologies, network scalability, node mobility, energy consumption etc, is a big challenge and requires designing new robust, adaptive, dynamic, and configurable routing protocols. With the arrival of the 5G and the future Internet, the latency time will be extremely reduced, this motivated us to propose a new protocol that entrust internet to transport a large part of control and data traffic of the network. In this paper, a dynamic multi-sink routing protocol (DMS-RP) is proposed for self-organizing multi-hop WSNs, iMANETs and IoT, it uses a sink-based auto-clustering of the network without using a dedicated clustering algorithm and it allows a multipoint-to-multipoint communication for local areas as well as for distant separated areas. Three multi-hop forwarding techniques of the proposed protocol are studied: B A S I C , M U L T I − P A R E N T and M U L T I − P A R E N T + S N R . We have implemented the proposed protocol as a new module for NS-3 simulator. The simulation results show that the proposed protocol achieves a largely better performance in terms of packet delivery, network delay, overhead, energy and network lifetime when compared with one of common multi-hop self-organizing routing protocols (AODV). Furthermore, the protocol analysis shows its high efficiency regarding scalability for static as well as for mobile networks where the studied performances are not significantly affected by the network size which makes it a suitable routing protocol for large-scale network architectures.

This paper presents a new hybrid approach for 3D face verification based on tensor representation in the presence of illuminations, expressions and occlusion variations. Depth face images are divided into sub-region and the Multi-Scale Local Binarised Statistical Image Features (MSBSIF) histogram are extracted from each sub-region and arranged as a third order tensor. Furthermore, to reduce the dimensionality of this tensor data, we use a novel hybrid approach based on two steps of dimensionality reduction multilinear and non-linear. Firstly, Multilinear Principal Component Analysis (MPCA) is used. MPCA projects the input tensor in a new lower subspace in which the dimension of each tensor mode is reduced. After that, the non-linear Exponential Discriminant Analysis (EDA) is used to discriminate the faces of different persons. Finally, the matching is performed using distance measurement. The proposed approach (MPCA+EDA) has been tested on the challenging face database Bosporus 3D and the experimental results show that our method achieves a high verification performance compared with the state of the art.

Channel characterisation for indoor visible-light communication systems is revisited. The purpose of this study is to evaluate the channel frequency selectivity, or in other words, the significance of inter-symbol interference (ISI) at the receiver and the necessity of channel equalisation to recover the transmitted data. The authors focus on the effect of the indoor channel by assuming no bandwidth constraint on the light-emitting-diodes and by considering a simple intensity modulation technique, excluding discrete multi-tone modulation. The channel impulse response (CIR) is first simulated using the iterative site-based method. Then, conventional metrics for evaluating channel frequency selectivity, that is, the root-mean-square delay spread and channel frequency response are investigated and their practical interest is discussed. The authors also consider the signal-to-ISI ratio (SIR), which they define based on the sampled (i.e. discrete-time) signal at the receiver, and demonstrate its usefulness in determining the necessity of channel equalisation at the receiver. They consider several link scenarios in a medium-size and a large room, and show the significance of the LOS components of CIR in determining the channel frequency selectivity. They also discuss the choice of the receiver filter and explain how it affects the SIR.

Positron Emission Tomography (PET) imaging is of importance for diagnosing neurodegenerative diseases like Alzheimer Disease (AD). Computer-aided diagnosis methods could process and analyze quantitatively these images, in order to better characterize and extract meaningful information for medical diagnosis. This paper presents a novel computer-aided diagnosis technique for brain PET images classification in the case of AD. Brain images are first segmented into 116 Regions Of Interest (ROI) using an atlas. Computing some statistical features (mean, standard-deviation, skewness, kurtosis and entropy) on these regions' histogram, we define a Separation Power Factor (SPF) associated to each region. This factor quantifies the ability of each region to separate AD from Healthy Control (HC) brain images. Ranking selected regions according to their SPF and inputting them to a Support Vector Machine (SVM) classifier, yields the same or slightly better classification results than when inputting the whole brain voxels or the 116 regions, but with less computational time. Comparison is also done using Random Forest classifier.

International audience; We propose a feature fusion method for crowd counting. By image feature extraction and texture feature analysis methods, data obtained from multiple sources are used to count the crowd. We count people in high density static images. Most of the existed people counting methods only work in small areas, such as office corridors, parks, subways and so on. Our method uses only static images to estimate the count in high density images (hundreds or even thousands of people), for example, large concerts, National Day parade. At this scale, we can't rely on only one set of features for counting estimation. Therefore, we use multiple sources of information, namely, HOG and LBP. These sources provide separate estimates and other combinations of statistical measurements. Using the support vector machine (SVM) classification technique, and regression analysis, we count the crowd with high density. The method gives good results in crowded scenes.

In the last few years, multilinear subspace learning has attracted large interest for dimensionality reduction and classification of multidimentional data. In this paper, we propose a new multilinear supervised projection method, called Multilinear Enhanced Fisher Discriminant Analysis (MEFDA), which improves the generalization capacity by decomposing the discriminate analysis process into a simultaneous diagonalization of the inter-class and intra-class scatter matrices resulting from the unfolding of tensor data. MEFDA uses the discriminant tensor criterion based on iterative procedure in order to obtain the optimal projection matrix in each k mode of tensor data. One virtue of MEFDA is that it can avoid the disadvantage of large dimensionality, in which the computational cost is reduced. Since the data encoded using high-order tensor, MEFDA respects the original structure and the natural geometry of face data. Therefore, the natural correlation of multi-dimension face data can be exploited. The proposed algorithm has been evaluated on three public 3D face datasets, FRGC V2.0, CASIA 3D and Bosphorus under the hard challenges of poses, expressions, illuminations and occlusions. The experimental results demonstrate that the proposed method significantly outperforms the current state of the art methods, with 99% and 97.02% at 0.001 False Acceptance Rate (FAR) for FRGC v2.0, and Bosphorus datasets, respectively, and 99.36% at evaluation Equal Error Rate (EER) for CASIA 3D dataset.

International audience; The purpose of this paper is to present a complete experimentation of the two failure modes in competition that can appear during short-circuit (SC) fault operation of single-chip 1,2kV SiC MOSFETs from different manufacturers including planar and trench-gate structures, well-known or recent devices. Ruggedness and selective failure modes are identified in relation with the power density dissipated by the chip and the simulated 1D-thermal junction. Finally, the chips of the devices which failed in a "fail-to-open" mode have been studied in order to find the physical reasons of this original and unusual fail-safe mode. Preferred presentation: [X] Oral [ ] Poster [ ] No preference Preferred track (please, tick one or number 1 to 3 tracks in order of preference: 1 = most suiting, 3 = least suiting) [ ] A-Quality and Reliability Assessment Techniques and Methods for Devices and Systems [ ]

This article presents the development of an intelligent technique of Adaptive-Neuro-Fuzzy Inference System (ANFIS) based on Maximum Power Point Tracking (ANFIS-MPPT) algorithm with PI controller in order to increase the performances of the photovoltaic panel system below change atmospheric circumstances. In this work, the mathematical principles of the ANFIS method were presented and developed using the software Matlab/Simulink. Moreover, the effectiveness of this ANFIS-MPPT technique is demonstrated by a comparison of the obtained results with others obtained from a classical (Perturb & Observe) P & O-MPPT method.From the analysis of the obtained results, the ANFIS-MPPT command provide better performances, respectable dynamic operations, quicker convergence and fewer fluctuations of working point nearby MPP compared to the classical P & O-MPPT under every irradiance conditions. We have confirmed, here, the capacity of the ANFIS-MPPT technique to increase not only the performance but also the tracking accuracy, speed and system stability under varying climatic conditions. Such improvement would be very beneficial when looking at the significance of the network coupled PV system.

International audience; With the current state-of-the-art computer aided manufacturing tools, the spatial resolution of hyperspectral sensors is becoming increasingly higher thus making it easy to obtain much more detailed information of the scene captured. However, the improvement of the spatial resolution also brings new challenging problems to address with signal dependent photon noise being one of them. Unlike the signal independent thermal noise, the variance of photon noise is dependent on the signal, therefore many denoising methods developed for the stationary noise cannot be applied directly to the photon noise. To make things worse, both photon and thermal noise coexist in the captured hyperspectral image (HSI), thus making it more difficult to whiten noise. In this paper, we propose a new denoising framework to cope with signal dependent nonwhite noise (SDNW), Pre-estimate—Whitening—Post-estimate (PWP) loop, to reduce both photon and thermal noise in HSI. Previously, we proposed a method based on multidimensional wavelet packet transform and multi-way Wiener filter which performs both white noise and spectral dimensionality reduction, referred to as MWPT-MWF, which was restricted to white noise. We get inspired from this MWPT-MWF to develop a new iterative method for reducing photon and thermal noise. Firstly, the hyperspectral noise parameters estimation (HYNPE) algorithm is used to estimate the noise parameters, the SD noise is converted to an additive white Gaussian noise by pre-whitening procedure and then the whitened HSI is denoised by the proposed method SDNW-MWPT-MWF. As comparative experiments, the Multiple Linear Regression (MLR) based denoising method and tensor-based Multiway Wiener Filter (MWF) are also used in the denoising framework. An HSI captured by Reflective Optics System Imaging Spectrometer (ROSIS) is used in the experiments and the denoising performances are assessed from various aspects: the noise whitening performance, the Signal-to-Noise Ratio (SNR), and the classification performance. The results on the real-world airborne hyperspectral image HYDICE (Hyperspectral Digital Imagery Collection Experiment) are also presented and analyzed. These experiments show that it is worth taking into account noise signal-dependency hypothesis for processing HYDICE and ROSIS HSIs.

Grey wolf optimizer (GWO) is a bioinspired iterative optimization algorithm which simulates the hunting process of a wolf pack guided by three leaders. In this paper, a novel discrete GWO is proposed: a random leader selection is performed, and the probability for the main leader to be selected increases at the detriment of the other leaders across iterations. The proposed discrete GWO is compared to another discrete version of GWO, using standard test functions.

International audience; Wireless visual sensor networks (WVSNs) are commonly using information technologies of modern networking and computing platforms. Today, visual network computing applications are faced with high demand of powerful network functionalities and performances. This paper proposes the specific WVSN nodes that are able to sense surrounding signals coming from the patient in the rehabilitation room and perform as local computations wirelessly communicated within the considered WVSNs. Assuming the WVSN platform for the rehabilitation supervision of patients, this paper discusses the specifications of the concept and the development of the WVSN nodes relying on KINECT and Raspberry Pi 3 (RPi 3) boards. Several technologies, such as RPi 3, Kinect, and ID sensor, are utilized in the proposed platform to realize sensors, image-capturing units, and processing cores. This collaborative platform consists of three nodes that are designed in three processing cores using the entire workflow to support the development steps. The communication of the elements between one to another and between the components takes place via Wi-Fi. The implementation of a preprocessing stage for component integration and the development of software development kit tools are necessary. In its final part, this paper summarizes the outlines of the proposed platform in different scenarios which run on nodes and also provides evaluation results that prove the feasibility of the approach.

Network-on-chip (NoC) is a promising communication medium for multi-processor SoCs. However, the increasing demand of flexible use-cases in SoC applications triggers researchers to focus on adaptive and more flexible NoCs. A dynamically reconfigurable NoC has been introduced in the last decade to provide more flexible and adaptive communication during run-time. Most of the previous works share a general title: dynamically reconfigurable NoC while their proposed reconfigurability focuses on a specific infrastructure and protocols. Thus, in this paper, we have presented an overview of state-of-the-art dynamically reconfigurable NoC by classifying them into a dynamically reconfigurable topology, routers, links, routing and switching techniques.

This paper presents an efficient framework for verification using 3D information based on high order tensor representation in uncontrolled conditions. The 3D depth images are subdivided into sub-blocks and the Multi-Scale Local Binarised Statistical Image Features (MSBSIF) + Multi-Scale local phase quantization (MSLPQ) histograms are extracted and concatenated from each block and organized as a 3rd order tensor. Moreover, two steps of dimensionally reduction to the face tensor are used. Firstly, Multilinear Principal Component Analysis (MPCA) is used to project the face tensor in a new subspace features in which the dimension of each mode tensor is reduced. After that, Enhanced Fisher Model (EFM) is applied to discriminate the faces of diverse persons in the database. Finally, the corresponding is achieved based distance measurement. The proposed approach (MPCA+EFM) has been evaluated on the challenging face database Bosporus 3D. The experimental results demonstrate that our method attains a high authentication performance.

International audience; In this paper, we present a Model Driven Engineering (MDE) methodology for facilitating the modeling of the partial reconfiguration process, and for implementing Dynamic Reconfigurable System-on-Chip (DRSoC). The rationale for this approach is to provide a modeling front-end that enables to visually compose a hardware platform, containing heterogeneous components, using both static and context-management hardware wrappers. A model transformations engine (MTE) processes the high-level models to obtain the inputs for the Xilinx dynamic partial reconfiguration (DPR) design flow, with the benefit of better exploiting and reusing the designer intentions regarding the allocation of tasks into reconfigurable areas. Furthermore, the automatic synthesis of a reconfiguration controller (RecOS) with context-management and task relocation capabilities is supported in this version of our tool. The latter feature is possible due to the integration of relocation tool OORBIT into the design chain, but we point out at other avenues of research. We present a case study in which we assess the benefits of the methodology and present a thorough analysis of the reconfiguration costs associated with the context and relocation management, showing speedups of 1.5x over other solutions.

Most of acoustical experiments face multipath propagation issues. The times of arrival of different ray paths on a sensor can be very close. To estimate them, high resolution algorithms have been developed. The main drawback of these methods is their need of a full rank spectral matrix of the signals. The frequential smoothing technique overcomes this issue by dividing the received signal spectrum into several overlapping sub-bands. This division yields a transfer matrix that may suffer rank deficiency. In this paper, a new criterion to optimally choose the sub-band frequencies is proposed. Encouraging results were obtained on real-world data.

Nowadays, H.264/AVC reaches very high resolutions, flexible depending on the application, which particularly leads to a very large amount of data compression. The motion estimation defines the most-complex process and the most interesting at the same time; its complexity appears with a consumption of more than 60% of total encoding time which make it difficult to achieve real-time encoding required in many video applications, but the interesting part is the flexibility to make its implementation variant according to the application. For this purpose, the following paper presents a speed real-time hardware implementation of integer motion estimation used in H.264/AVC, the method is based on the full-search block-matching algorithm. The proposed architecture calculates best motion vector using a parallel process. It is composed of three processor modules and a comparator three values. Implementation results based on Field Programmable Gate Array devices uses Xilinx Virtex7 XC7VX550T show performance...

We consider channel estimation in amplify-and-forward wireless relay networks when non-orthogonal coherent distributed space-time coding is done at relays. We propose to perform iterative channel estimation and data detection based on parallel interference cancelation (PIC) at the receiver. The schemes that we propose are based on soft or hard estimates of the transmitted data symbols. We illustrate the performance improvement compared to the classical approach of pilot-only-based channel estimation and minimum mean-square-error detection. Also, we show that semi-blind channel estimation together with PIC detection is particularly promising for large signal constellations. Moreover, we show that the use of hard estimates of symbols in the proposed iterative receiver provides satisfying results, while requiring a lower computational complexity.

Target detection is a key issue in processing hyperspectral images (HSIs). Spectral-identification-based algorithms are sensitive to spectral variability and noise in acquisition. In most cases, both the target spatial distributions and the spectral signatures are unknown, so each pixel is separately tested and appears as a target when it significantly differs from the background. In this paper, we propose two algorithms to improve the signal-to-noise ratio (SNR) of hyperspectral data, leading to detectors that are robust to noise. These algorithms consist in integrating adaptive spatial/spectral filtering into the adaptive matched filter and adaptive coherence estimator. Considering the HSIs as tensor data, our approach introduces a data representation involving multidimensional processing. It combines the advantages of spatial and spectral information using an alternating least square algorithm. To estimate the signal subspace dimension in each spatial mode, we extend the Akaike information criterion, and we develop an iterative algorithm for spectral-mode rank estimation. We demonstrate the interest of integrating the quadtree decomposition to perform an adaptive 3-D filtering and thereby preserve the local image characteristics. This leads to a significant improvement in terms of denoised tensor SNR and, consequently, in terms of detection probability. The performance of our method is exemplified using simulated and real-world HYperspectral Digital Imagery Collection Experiment images.

In the hyperspectral images (HSI) acquired by the new-generation hyperspectral sensors the signal dependent noise is an important limitation to the detection. Therefore, noise reduction is an important preprocessing step to analyze the information in the hyperspectral image. A signal dependent noise cannot be reduced by conventional linear filtering. Therefore, a new method based on Parallel factor analysis (PARAFAC) decomposition is proposed to estimate the noise of hyperspectral remote sensing image. Then, the estimated noise is used for whitening the colored structural noise. By using this transformation, the data noise from new-generation hyperspectral sensors is diminished, thereby allowing a minimization of negative impacts on hyperspectral detection applications.

Sources localization is an important theme of antenna and signal processing researches since decades. It has a lot of applications for source characterization, which leads to object detection, used in submarine field as well as seismic and remote sensing studies. The particularity of array processing is that we are looking at spatio-temporal vector signals which are generated by both radiating sources or reflected by buried objects and ambient medium. These signals are generally corrupted by an additive noise. When the number of sources is greater than the number of the sensors of the smart antenna system, classical methods become unusable. Recently, a new subspace method called incoherent sensor by sensor processing (ISSP) for estimation of directions of arrival (DOAs) is proposed. It is based on a study sensor by sensor to estimate firstly times of arrival(TOAs), then the DOAs of the sources are estimated. The main drawback of this method is the computational load needed to estimate accurately the DOAs. In this paper, we propose new faster methods thanks to the numerical complexity reduction of the existing algorithms. This last point is important nowadays to make embedded systems faster.

Linear structures such as blood vessels in medical images are important features for computer-aided diagnosis and follow-up of many diseases. In this letter a new tracking-based segmentation method is proposed to detect blood vessels in retinal images. Bayesian segmentation with the Maximum a posteriori (MAP) Probability criterion is used for that purpose. Tests on simulated and retinal images are presented and compared with a vessel detection technique. Our method performs better results.

International audience; Since the transmission of the uncompressed image in the context of wireless visual sensor networks (WVSNs) consumes less energy than transmitting the compressed image, developing energy-aware compression algorithms are mandatory to extend the camera node's lifetime and thereby the whole network lifetime. The present paper studies a low-complexity image compression algorithm in the context of WVSNs. This algorithm consists of applying a pruning approach on a DCT approximation transform. The scheme is investigated in terms of computation cycles, processing time, energy consumption and image quality. Experimental works are conducted using the Atmel Atmega128 processor of Mica2 and MicaZ sensor boards. Simulation results show that the studied scheme can exhibit a competitive performance when compared against other algorithms. Furthermore, the scheme can achieve the best tradeoff between energy consumption and image quality.

International audience

Visible light communications (VLCs) have received extensive attention in the research community thanks to their advantages of high bandwidth, low cost, robustness to electromagnetic interference, operation in an unregulated spectrum, and high degree of spatial confinement in indoor scenarios. One of the main limitations for high data-rate transmission in VLC systems is the limited modulation bandwidth of commercial light emitting diodes. To circumvent this limitation, spectrally efficient modulation schemes should be used. Optical orthogonal frequency-division multiplexing (O-OFDM)-based schemes have become very popular and several proofs of concept have shown their ability to attain over gigabits per second transmission rates. We consider here the use of pulse amplitude modulation and carrier-less amplitude and phase modulation schemes together with frequency-domain equalization (FDE) at the receiver as interesting alternatives to O-OFDM. We show the advantages of the former schemes in terms of the peak-to-average-power ratio, and demonstrate through numerical results the merits of FDE-based signaling in attaining high data rates.

The integration of virtualized FPGA-based hardware accelerators in a cloud computing is progressing from time to time. As the FPGA has limited resources, the dynamic partial reconfiguration capability of the FPGA is considered to share resources among different virtualized IPs during runtime. On the other hand, the NoC is a promising solution for communication among virtualized FPGA-based IPs. However, not all the virtualized regions of the FPGA will be active all the time. When there is no demand for virtualized IPs, the virtualized regions are loaded with blank bitstreams to save power. However, keeping active the idle components of the NoC connecting with the idle virtualized regions is not optimal for the total power consumption. In this paper, we have proposed a runtime scalable NoC where the NoC size is scaled up and down on demand. The experimental result shows that the total power consumption of the NoC can be reduced by more than 15% if the runtime scalable Noc is used.

This paper firstly describes the development and evaluation of a project course which yields university students building a complete hardware and software chain for a digital image processing application. To arouse the interest and learning initiative of students, we propose them to build a setup including a Raspberry Pi® and image processing programmes. This inexpensive single board computer answers today's issues in energy saving and permits to review fundamental hardware and software principles. Secondly, we propose a low-cost setup for a time-limited practical work: a Raspberry Pi® is shared and controlled remotely by several student pairs: we emphasize the interest of collaborative work, and we provide knowledge and skills about micro-computers to a large number of students simultaneously.

International audience; Mobile ad hoc networks (MANETs) are becoming an emerging technology that offer several advantages to users in terms of cost and ease of use. A MANET is a collection of mobile nodes connected by wireless links that form a temporary network topology that operates without a base station and centralized administration. Routing is a method through which information is forwarded from a transmitter to a specific recipient. Routing is a strategy that guarantees, at any time, the connection between any two nodes in a network. In this work, we propose a novel routing protocol inspired by the cuckoo search method. Our routing protocol is implemented using Network simulator 2. We chose Random WayPoint model as our mobility model. To validate our work, we opted for the comparison with the routing protocol ad hoc on-demand distance vector, destination sequence distance vector and the bio-inspired routing protocol AntHocNet in terms of the quality of service parameters: packet delivery ratio and end-to-end delay (E2ED).