Search

Electroencephalography (EEG) signal recorded during motor imagery (MI) has been frequently used in noninvasive Brain-Computer Interface (BCI) is a new type of device that allows direct communication between user's brain and machine. This paper proposes a novel solution for extraction and classification of left/right hand, both feet, and tongue movement by exploiting two approaches of deep learning such as artificial neural network ANN and convolutional neural network CNN. A wide range of spatial and frequency domain features are extracted from the EEG signals and to train an ANN and CNN networks to perform the classification tasks. The EEG signals of mental tasks are extracted and classified by these architectures. In addition, the proposed methods are validated by the EEG dataset of the BCI competition IV-2a and we compared them with each other. The results show that the CNN model surpasses the ANN model by an accuracy value of 60.55%.

In this study, three feedstocks were studied (Grape marc, Kenaf stems, flax shive) for biochar production. The pyrolysis operation was carried out at 400; 500 and 600°C and a constant heating gradient rate of 5 °C/min for a future potential use in agriculture. The experimental results showed that the biochar yields decreased with the increase of the applied pyrolysis temperature and were approximately constant above 500°C for all feedstock .The physico-chemical properties of grape marc and Kenaf indicated that they have the highest amount of potentially available minerals with an interesting value of surface area and microposity. These properties make these biochars suitable candidates for soil amendments as well as pollutants removal from aqueous and gaseous effluents.

The multidisciplinary collaboration can be efficiently supported by the Model-Based Systems Engineering (MBSE) approach. However, even if the MBSE approach presents a brilliant method of development of complete architectures, this approach does not avoid system defects during or after the development process. In this paper, we describe the benefits to integrate agile methods into MBSE approaches and we propose in particular to introduce the Agile Scrum method in a MBSE design methodology for multidisciplinary systems.

The recent popularity of deep neural networks (DNNs) has generated considerable research interest in performing DNN-related computation efficiently. However, the primary focus is usually very narrow and limited to (i) inference – i.e. how to efficiently execute already trained models and (ii) image classification networks as the primary benchmark for evaluation. Our primary goal in this work is to break this myopic view by (i) proposing a new benchmark suite for DNN training, called TBD1 TBD is short for Training Benchmark for DNNs, which comprises a representative set of eight DNN models and covers six major machine learning applications: image classification, machine translation, speech recognition, object detection, adversarial networks, reinforcement learning, and (ii) performing an extensive performance analysis of these models on three major deep learning frameworks (TensorFlow, MXNet, CNTK) across different hardware configurations (single-GPU, multi-GPU, and multi-machine). We present a new toolchain for performance analysis for these models that combines the targeted usage of existing performance analysis tools, careful selection of performance metrics, and methodologies to analyze the results. We also build a new set of tools for memory profiling in three major frameworks. These tools can shed light on precisely how much memory is consumed by different data structures (weights, activations, gradients, workspace) in DNN training. Using our tools and methodologies, we make several important observations and recommendations on where future DNN training research and optimization should be focused.1TBD is short for Training Benchmark for DNNs

Nowadays, cryptographic applications are becoming of paramount importance in order to guarantee ultimately secure communications. Performances of classical and quantum key distribution and encryption algorithms are strongly dependent on the used Random Number Generator (RNG). A good RNG must produce unpredictable, unreproducible and unbiased sequences of numbers. For this reason, many true random number generators relying on chaotic physical phenomena, such as chaotic oscillations of high-bandwidth lasers [1, 2] or polarization chaos from a VCSEL diode [3], have been developed. In this work, we propose a RNG implementation based on a different physical mechanism than the ones previously used in literature: the polarization chaos induced in an optical fiber by the nonlinear interaction between a forward beam and its amplified backward replica. Basically, we operate the device called Omnipolarizer, originally used in our previous works as an all-optical polarization attractor or beam splitter [4], in an additional operating mode, namely the chaotic mode [5].

An iris recognition requires parametric modeling texture. The extracted model should characterize the individual corresponding to considered iris. Such a model is often referred to as biometric signature. Several approaches to uniquely specify an iris by extracting parameters characteristic of its texture exist in the literature. An original approach based on an analysis by the Meyer wavelet of the iris texture is detailed in this paper. A comparative study between our approach and some techniques that have been studied, implemented and tested in subsequent work is carried out on the CASIA V.1 database. The experimental results show that the proposed method has promising performances.

The road safety is a problem which was approached by several countries following a big raise of the number of accidents. The drowsiness represents one among the causes of the road accidents. The accidents related to the drowsiness often occur on the highways, but also on the main roads, even inside the localities. Today, it is possible to detect the state of tiredness of the driver with the development of the technology of the computer vision. The results of research in physiology show that the first level of lack of vigilance appears by an increase in the frequency of the yawn. In this work, we propose a novel approach for yawning detection for monitoring driver fatigue. In fact, our approach rests on the study of the spatio-temporal descriptors of a nonstationary and non-linear signal. This approach is evaluated by both YawDD [1] and our MiraclHB [2] databases. The evaluation shows many promising results and shows the effectiveness of the suggested approach.

International audience; —Web applications have become one of the most popular targets of attacks during the last years. Therefore it is important to identify the vulnerabilities of such applications and to remove them to prevent potential attacks. This paper presents an approach that is aimed at the vulnerability assessment of Web applications following a black-box approach. The objective is to detect vulnerabilities in Web applications and their dependencies and to generate attack scenarios that reflect such dependencies. Our approach aims to move a step forward toward the automation of this process. The paper presents the main concepts behind the proposed approach and an example that illustrates the main steps of the algorithm leading to the identification of the vulnerabilities of a Web application and their dependencies.

RRABIDS (Ruby on Rails Anomaly Based Intrusion Detection System) is an application level intrusion detection system for applications implemented with the Ruby on Rails framework. It is aimed at detecting attacks against data in the context of web applications. This anomaly based IDS focuses on the modeling of the application profile in the absence of attacks (called normal profile) using invariants. These invariants are discovered during a learning phase. Then, they are used to instrument the web application at source code level, so that a deviation from the normal profile can be detected at run-time. This paper illustrates on simple examples how the approach detects well known categories of web attacks that involve a state violation of the application, such as SQL injections. Finally, an assessment phase is performed to evaluate the accuracy of the detection provided by the proposed approach.

The performance of a quantum-dash Fabry-Perot laser is experimentally characterized for 40 Gb/s RZ-DPSK and NRZ-DPSK clock recovery, 40 Gb/s RZ-OOK all-optical 3R regeneration, and 160 Gb/s OTDM sub-harmonic clock recovery.

Optimization of novel InAs/InP quantum dash nanostructures has allowed the realization of mode locked lasers that exhibit unprecedented performance, enabling subpicosecond pulse generation at ≫300 GHz repetition rates and very low timing jitter.

We demonstrate experimentally low-timing-jitter all-optical clock recovery for 40-Gb/s NRZ-DPSK signals using a mode-locked quantum-dot Fabry-Perot laser. The clock recovery performance shows a high tolerance to optical noise and signal waveform distortion.

Semantic Web ontologies are being increasingly used in modern text analytics applications and ontology-based information extraction (OBIE) as a means to provide a semantic backbone either for modelling the internal conceptual data structures of the text analytics (TA) engine or to model the knowledge base to drive the analysis of unstructured information in raw text and subsequent Knowledge acquisition and population. creating and targeting language resources (LR)s from a TA to an ontology can be time consuming and costly.The authors describe a user-friendly method for ontology engineers to augment an ontologies with a lexical layer which provides a flexible framework to identify term mentions of ontology concepts in raw text. In this paper we explore multilinguality in these lexical layers using the same framework. We discuss a number of potential issues for the ldquolinguistic lightrdquo lexical extensions for ontologies (LEON) approach when looking at languages more morphologically rich and which have more complex linguistic constraints than English. We show how the LEON approach can cope with these phenomena once the morphological normaliser used in the lexical analysis process is able to generalise sufficiently well for the language concerned.

Dispersion impact on electrical signal generation at 54.8 GHz of a quantum dash (QDs) mode-locked Fabry-Perot (FP) lasers emitting at 1.55 mum has been measured. Optimum fibre length values that correspond to dispersion compensation have been found. For these optimal conditions a pulse width of 480 fs has been measured. These optimum fibre length values have been applied to the integration of two architectures of phase noise reduction schemes: a coupled opto-electronic oscillator (COEO) with a single self-injection loop associated to a self-phase-locked loop (PLL) and a dual loop COEO. A phase noise reduction of more than 30 dB at an offset frequency of 20 kHz has been obtained for the second architecture.

We demonstrate ~860 fs pulse generation at a 21.31 GHz repetition rate using a two-section single quantum well mode-locked laser emitting at 1.54 mum. A very low 30 kHz linewidth of the radio frequency (RF) spectrum is reported.

We demonstrate ~590 fs pulse generation at a 245 GHz repetition rate using a one-section Fabry-Perot quantum-dash-based laser. A time-bandwidth product of 0.41 and average output power in excess of 20 mW are achieved.

Electrical spectrum line-width is reduced in mode-locked FP 1.55 μm-QD laser diode through optical confinement factor optimization. From optimized structures, we obtained nearly Fourier-transform limited pulses at 10 GHz, with an averaged width of 8 ps over 10 nm.

Error-free transmission over 50 km single mode fiber of 8 WDM ITU channels with 100 GHz channel spacing is demonstrated at 10 Gbit/s by using comb generation in a quantum dash mode-locked laser.

2.2 ps pulses at 42.72 GHz after spectrum filtering and 1.1 ps pulses after chirp compensation are generated using mode-locked laser diodes based on 1.55 mum quantum dash.

This paper presents the influence of the process of migration between populations in GENO-FLOU, which is an environment of learning of fuzzy knowledge bases by genetic algorithms. Initially the algorithm did not use the process of migration. For the learning, the algorithm uses a hybrid coding, binary for the base of rules and real for the data base. This hybrid coding used with a set of specialized operators of reproduction proven to be an effective environment of learning. Simulations were made in this environment by adding a process of migration. While varying the number of populations, the number of generations and the rate of migration or simply the migration of the best elements, on various types of problems. In general, simulations show a significant improvement of the results obtained with migration. The variation of these parameters makes it possible to conclude on the dominating importance of the number of migrant generations.

Summary form only given. Willingness to customize microprocessors for specific applications has uniquely positioned IBM as the microprocessor supplier of choice to the games market. An identical approach of customizing microprocessors for network processing provides tremendous processing power and gives much more flexible networking functions support compared to what is offered by classical network processors with hardware co-processors. Evolution of the gaming industry is driving a need for microprocessors whose structure supports both gaming and networking functions with a very high level of programming flexibility with a consistent programming model, as well as the tremendous performance required by on-line gaming on networking

The key characteristics of network processing are performance, ease-of-use, and flexibility. As technology continues to become increasingly more powerful, it becomes possible do networking functions using a nonspecific processor rather than developing a specific design tuned specifically for networking requirements. Also, as the industry continues to drive for increasing performance, we are now realizing that limitations on continual frequency-increases and therefore others ways to improve performance are now being considered. Such features as multiprocessing and multithreading are now making their ways into the architectures of the nonspecific processor to improve performance by reducing external latencies. With this trend in place, we are now at a time where many standard nonspecific architectures meet the capabilities needed for network-processor-based implementations. To highlight the many advantages of choosing to use these nonspecific new design architectures in the networking segment, we need to address the technical challenges associated with meeting the performance requirements of today's high performance network processors.

Electrostatic discharge (ESD) protection in silicon-on-insulator (SOI) semiconductor technology is perceived as a major roadblock for the SOI technology to become a viable mainstream contender for high-performance advanced CMOS semiconductor chips (Hu, 1994; Colinge, 1991). In this paper, our results in four successive SOI technology generations demonstrate that excellent ESD protection levels are achievable in SOI chips with no additional masking steps, process implants, costs or ESD design area. ESD results also show that the ESD robustness of the SOI ESD device is improving with partially depleted SOI MOSFET scaling from 0.25 to 0.12 /spl mu/m L/sub eff/ technology generations (Shahidi et al., 1999; Voldman et al., 1995, 1997, 1999). By allowing the ESD network minimum design to scale with the technology, improved ESD results are evident in each generation with no indication of any SOI-specific ESD limitations. For future technology generations below 1.5 V V/sub DD/ power supply, continued improvement is anticipated due to buried-oxide scaling, lower trigger voltages, dynamic threshold voltage MOSFET (DTMOS) techniques and ESD I/O design learning (Voldman et al, 1997; Assaderaghi et al., 1994). ESD protection of partially depleted SOI technology is not a problem or technology concern using our proposed ESD methodology.

A RISC microprocessor remapped in SOI technology exploits the advantages of SOI to boost processor frequency by 20% to 580MHz at 2.0V and 85/spl deg/C and fast process. The separation by implanted oxygen (SIMOX) SOI process produces partially-depleted devices. Source and drain capacitances are reduced by an order of magnitude, improving gate delay by 12%. Reduction in body-bias effects on device stacks and passgate topologies results in an additional 15%-25% improvement. Speed gains of up to 35% are achieved in some designs. The frequency-limiting paths in this processor are dominated by SRAM access and self-timed dynamic circuits whose timing had to be relaxed to guarantee functionality.