ISPA'19: THE SIXTH INTERNATIONAL CONFERENCE ON THE IMAGE AND SIGNAL PROCESSING AND THEIR APPLICATIONS
PROGRAM FOR SUNDAY, NOVEMBER 24TH
Days:
next day
all days

View: session overviewtalk overview

09:30-10:00Coffee Break
10:00-12:30 Session 4A: Image Processing
10:00
Fusing Palmprint,Finger-knuckle-print for Bi-modal Recognition System Based on LBP and BSIF
PRESENTER: Bilal Attallah

ABSTRACT. Multimodal biometrics is an evolving technology in the fields of security. Biometrics system reduces the effort of remember a memorable password. Multimodal biometrics system uses two or more traits for efficient recognition. This paper presents a hand biometric system by fusing information of palmprint and finger knuckle. To this end, BSIF ( Binarized Statistical Image Features) filter and LBP (Local binary patterns) coefficients are employed to obtain the Finger-knuckle-print and palm-print traits, and subsequently selection of the features vector is conducted with PCA (Principal Component Analysis) transforms in higher coefficients. To match the finger knuckle or palm-print feature vector, the (ELM) Extreme learning machine is applied. According to the experiment outcomes, the proposed system not only has a significantly high recognition rate but it also affords greater security compared to the single biometric system.

10:20
Roads Recognition using Mathematical Morphology

ABSTRACT. In this paper, a road extraction technique from very high resolution remote sensing images is proposed based on mathematical morphology. The approach is efficient so that it is unaffected by road shapes. However, it has some drawbacks that slightly reduce the performance, such as sensitivity to shadows (over-crossing roads) and other objects that coexist with roads (e.g. trees on road edges). The task starts by extracting some features from an image (RGB or gray-scale), then moving to the classification process using the SVM classifier, finally a filtering stage is considered for results refinement.

10:40
Performances Improvement of Chaos-Based Robust Digital Image Watermarking using Advanced Techniques
PRESENTER: Redouane Kaibou

ABSTRACT. This paper presents a research study leading to a robust chaos-based scheme for digital image watermarking applications in the transform domain. The steps undergone include combining, in a repeating process, three performant conventional mathematical transforms of Direct Cosine Transform (DCT), Discrete Wavelet Transform (DWT) and variants with three different state-of-the-art advanced techniques namely Discrete Curvelet Transform (DCuT), Discrete Radon Transform (DRaT) and Finite Ridgelet Transform (FRT). Results have shown a neat improvements in watermarking imperceptibility and robustness each time especially after including Singular Value Decomposition (SVD) which has enhanced embedding capability as well. Methods evaluation has followed an exhaustive series of perfomed tests that applied a range of evaluation metrics for imperceptibility. In addition, Robustness of the final scheme has also been confirmed by applying a variety of attacks on the resulting images. Finally, security aspect has also been improved after a comparative implementations of popular four chaotic models namely Arnold Cat Map (ACM), Lorenz Cat Map (LCM), Rossler Cat Map (RCM) and Genesio-Tesi Cat Map (GTCM) that producing a secure image encryption levels. All Implementations have been validated using a sufficent number of color image from a large dataset.

11:00
A connected component-based approach for text-independent writer identification

ABSTRACT. Writer identification from off-line images of handwriting is a challenging task. In this work we assess the performance of textural descriptors for writer identification on different writing styles. The proposed method is different from the existing texture based methods: the earlier methods extract texture information at page, paragraph, fragment level to get a document descriptor, while the proposed method exploits the texture at a connected-component level. Specifically, an improvement of the textural features, the contour direction, angle, and length distributions are explored. Using these texture descriptors, the occurrence histograms are calculated in order to determine the similarities between different images. For the evaluation, the IFN/ENIT (411 writers) and IAM (657 writers) datasets were used. In our characterization of the individual and combination of textural features of connected-components we show that the proposed method outperforms the state-of-the-art algorithms and archives the best performance.

11:20
Weber pattern and Binarized Statistical Image Features encoded CNN for fingerprint liveness detection

ABSTRACT. This paper presents a new fingerprint liveness detection system based on local features Simplified Weber Local Descriptor (SWLD) and Binarized Statistical Image Features (BSIF) as a local textural descriptor encoded CNN models. SWLD feature and BSIF feature vector are combined together to ensure the preservation of the local intensity information and the orientation gradient in WLD is substituted by a recent Binarized Statistical Image Features (BSIF). After an appropriate feature selection, a trained linear-kernel SVM classifier makes the final live/fake decision. Our experimental analysis on two publicly available databases LivDet2011 and LivDet2013, comprising datasets collected from various sensors, show that the classification performance of our proposed method is effective and meanwhile, provides a better detection accuracy compared with the majority of previously published results, which are or not based on the binary pattern descriptors.

11:40
An Improvement of Spatial Fuzzy C-means Clustering Method for Noisy Medical Image Analysis
PRESENTER: Beddad Boucif

ABSTRACT. Medical image Segmentation plays a major role in MRI images processing; it’s performed before the analysis and decision-making stages in several medical processes. Many investigators have developed several Fuzzy C-means methods. In this work, a reliable automatic segmentation algorithm based on the spatial FCM clustering is developed to minimize the effect of noise and intensity inhomogeneities. This approach combines two properties of the Spatial FCM using neighbor's statistical characteristics and pillar k-means. The proposed system has been implemented with simulink. Experimental results on brain MRI images show the improvement and increase the segmentation accuracy.The Current research work is compared with some well-known existed methods to show the effectiveness that contributes to the development of tools needed in computer aided diagnosis systems aiming to assist specialists in making diagnosis decisions.

12:00
Land Cover Change Detection based on Homology Theory
PRESENTER: Meriem Khelladi

ABSTRACT. We exploit homology theory (HT) for land cover change detection in remote sensing. Shape disclosure is achieved through the Vietoris-Rips simplicial complex constructor. We propose to use persistent homology to overcome the land cover change detection problem in remote sensing. Shape disclosure is achieved through the Vietoris-Rips simplicial complex constructor. Qualitative Features (Betti numbers) are computed from a series of simplicial complexes at each level of resolution. Local structures are unfolded gradually throughout the observation sequence generated from a pixel in the input image at two different points in time. The weighted Bottleneck distance is used to detect the change satellite images. Experimental results demonstrate the efficiency of our methodology when compared to some major state-of-the-art approaches.

10:00-12:30 Session 4B: Telecommunication
10:00
Efficient Demultiplexing of Two Channels Based on Quasi-Square Ring Resonator for Integrated Optics

ABSTRACT. In this work a dual band demultiplexer was achieved based on a 2D photonic crystals with cylindrical gallium arsenic GaAs dielectric rods in square air lattice slab, the proposed architecture is based on ring resonators that consist of four nano-ring cavities embedded in a micro cavity, the demultiplexer functions around 1.330 and 1.550 µm simultaneously. The achieved results show a quality factor of 155 and a transmission efficiency of 85% and 95% around 1.330 and 1.550 µm respectively.

10:20
UWB Decoupling Approach for Two extremely Close Monopole Antennas
PRESENTER: Abderrahim Kouar

ABSTRACT. In this work, a new antenna system for UWB applications is proposed. Two identical monopoles are placed symmetrically on FR-4 substrate. The elements are extremely closed (spacing of about 0.13λ, center to center). The high mutual coupling between the radiating elements has been considerably reduced by exploiting a hybrid technique based on Defected Ground Structure (DGS) and parasitic elements (stub) integrated in the ground plane, hence isolation close to 20 dB is achieved. Besides, an ECC less than 0.01 is obtained which demonstrates good diversity performance. The antenna array is highly compact ( 29×27 mm 2 ), with wide operating frequency band (3.1 to 10.8 GHz). Further, acceptable agreement is demonstrated between simulations and measurements. Hence, the proposed system is recommended for recent wireless communications systems and UWB applications, thanks to its attractive performances, compactness and wide frequency range.

10:40
A Comparative Analysis of Various Zero Cross Correlation Codes in SAC-OCDMA Systems

ABSTRACT. The main purpose of this study is to present a comparison of different codes used for SAC-OCDMA systems. The main feature of compared codes is that they have a zero cross-correlation which eliminates both MAI and PIIN. Hence, a Direct Detection is used at a receiver, which reduces the complexity of receiver structure thereby reducing system cost. The comparison is based on some parameters: cardinality or number of users supported by each code, the effective received power from every user, the code length and BER. The results presented are based on the theoretical calculation of SAC-OCDMA systems using direct detection.

11:00
Hybrid Linear Precoding Strategy for Multiuser Massive MIMO Millimeter Wave Systems
PRESENTER: Baghdad Hadji

ABSTRACT. Multiuser-MIMO Millimeter wave based communication system has been considered as a current area of research for the future communication systems to serve multi-user simultaneously with a multiple data streams. Hybrid analog/digital architecture is the most satisfied architecture of the mmWave MIMO communication objectives and can be applied at both the base and mobile stations. Thus, exploiting the hybrid precoding/combining techniques offered by this architecture allow the realization of the multi-stream/multi-user transmission. We propose in this paper, a low computational complexity linear hybrid precoding schemes based on classical linear precoding and SVD decompostion of each channel user to avoid information exchange by the feedback where each user is powered by hybrid architecture. The simulation results demonstrate that our proposed two-stage hybrid linear precoding schemes can achieve a higher spectral efficiency compared with existing methods where each user is equipped with a single combiner and confirms its better management of the multi-user interference.

11:20
Study of mmWave channels for different scenarios

ABSTRACT. the fifth generation (5 G) technology is one of the most important technologies that will revolutionize the world of modern technology. The use of 5G will provide us with great advantages, including the high flow of up to 1 Gb/s, network capacity up to 1Million sub/km2 , the quality of service is high and sophisticated. One of the most important things on which the technology depends on the 5G is the high frequency and the socalled millimeter waves (mmWave) where the wavelength is limited between 1 mm and 10 mm and the equivalent of the range frequency between 30 and 300 GHz. They are greatly affected by air, conditions atmospheric, building and forest of our environment. In This paper, we will present a study of mmWave channels 38, 60 and 73 GHz for different scenarios using the New York University Simulator (NYUSIM) which isdeveloped by researchers at New York University (NYU).

11:40
Performance Comparison of Different SAC-OCDMA-FSO Detection Techniques in Presence of Atmospheric Losses
PRESENTER: Amel Aissaoui

ABSTRACT. The development of many high bit-rate multimedia applications has emphasized the demand for service differentiation or prioritization techniques to ensure end-user quality-of-service (QoS) necessities. This paper focuses on utilizing different detection schemes in spectral-amplitude coding optical code-division multiple-access (SAC-OCDMA) systems to support 'triple-play' services (voice, video, and data) with diverse QoS requirements. The system is simulated with enhanced double weight (EDW) code for complementary subtraction detection (CSD), modified-AND detection, as well as single photodiode detection (SPD) technique via OptiSystem software.

12:00
A New Conical Helix Antenna base on MEMS ‎Technology using Genetic Algorithms

ABSTRACT. The manufacturing of Micro-Electromechanical Systems ‎‎(MEMS) antennas has recently outperformed major hurdles by ‎reducing the long and expensive development cycles due to ‎adequate Computer Aided ‎Design (CAD) equipments with ‎advanced optimization methods to address the special need for a ‎new class of terahertz (THz) antennas, and ‎model accurately the ‎physical behavior of such three dimensional (3D) micro-devices. ‎In this paper a new design of a highly miniaturized MEMS ‎‎conical helix antenna modeled by High Frequency ‎Structure ‎Simulator (HFSS) based on Genetic Algorithms (GA), is ‎proposed for THz wireless systems to present very low ‎return ‎losses of ‎less than -20dB to -70dB and excellent voltage standing ‎wave ratios of less than 1.20 to 1.00 for a wide frequency band ‎ranging from 3.5 to 4.5 THz, resulting in excellent ‎‎electromagnetic response and good directivity due to the high ‎geometrical accuracy achieved using effective automatic ‎techniques to rectify various tunable parameters embedded in ‎‎silicon platform divided into helix variables and feeding line ‎‎characteristics.‎

12:30-13:30Lunch Break
14:15-15:15 Session 6: Poster Session I
Location: Poster Session
14:15
A comparative study of adaptive and non-adaptive IQ imbalance compensation methods in RX system

ABSTRACT. Different adaptive and non-adaptive compensation methods have been proposed in literature to compensate the problem of in phase and quadrature (IQ) imbalance in digital transmitter and receiver communication systems. In this paper, a comparative study is presented between an adaptive filtering compensation algorithm using the fast newton transversal filter algorithm combined with backward blind source separation (FB-FNTF) and a non-adaptive compensation method using the characteristics of I and Q signals in time domain, which have identical power and are uncorrelated . Their performance in terms of constellation diagram and symbol error rate (SER) are then evaluated for high-order quadrature modulation and high IQ imbalance values.

14:15
Improving Acoustic Echo Cancellation in Hands-free Communication Systems
PRESENTER: Mahfoud Hamidia

ABSTRACT. Acoustic echo cancellation (AEC) is the technique used to improve the speech quality in communication systems. This technique removes the undesirable acoustic echo generated by the acoustic coupling between the loudspeaker and the microphone of the terminal devices. In addition, AEC is generally based on adaptive filtering. This paper presents a new structure of AEC system to improve the performance of the conventional AEC. Simulation results using normalized least mean square (NLMS) algorithm demonstrate that better performance can be achieved by the proposed structure compared to the conventional structure of AEC. Therefore, three criteria measures are used to evaluate the proposed structure of AEC: misalignment, mean square error (MSE) and echo return loss enhancement (ERLE).

14:15
A NOVEL MULTIPATH CHANNEL ESTIMATION APPROACH FOR OFDM COMMUNICATION SYSTEMS

ABSTRACT. In this work, we propose a novel approach to estimate the parameters of the multipath channel to improve the performance of the wireless communication system using the OFDM modulation. The pilot pulse is inserted within the 0’th subcarrier of every OFDM symbol. A simple and efficient FFT algorithm is used to estimate the channel parameters from the frequency spectrum of the matched filter output of the pilot signal. The computer simulations show the capability of this approach to estimate the channel parameters even when the received signal is contaminated with the additive white Gaussian noise (AWGN). An important feature of the proposed algorithm is that the number of channel paths need not be known.

14:15
Dual Band Micro-strip Patch Antenna for RFID UHF far-field Reader Applications.

ABSTRACT. We propose in this paper a new design of RFID Dual Band reader micro-strip patch antenna. The design and analysis issues of RFID reader antenna for 2.4 GHz and 5.8 GHz UHF far-field application are presented. A dual band RFID Reader antenna is designed using RF4 substrate. This kind of substrate is characterized by a permittivity 4.3 and thickness 1.5 mm. We investigate a parametric study of slot in the patch to create a dual resonance. This modification shows the effect of the adaptations of two frequencies which are dedicated to the RFID band application. The obtained results are very encouraging regardless the return loss, the impedance matching antenna efficiency in 2.4 and 5.8 GHz bands, and 1.99 dBi and 3.5 dBi gain at frequency 2.4 and 5.8 GHz respectively.

14:15
Low Power Class D Audio Amplifier with High Performance and High Efficiency

ABSTRACT. Class D audio amplifiers are used in the audio devices. Nowadays, they are preferred due to their high power efficiency compared to the other classes such as: class A, B and AB. In this paper, a class D audio amplifier is designed. It uses MOSFETs in switching mode, a closed-loop with second order integrator, pulse width modulator and low pass filter. The system is simulated with Orcad/PSpice software. Simulated results show good values of THD+N, SNR and PSRR and Efficiency

14:15
CEEMDAN and fast kurtogram method for bearing fault detection.

ABSTRACT. A time-frequency analysis based on CEEMDAN and Fast kurtogram for bearing fault detection is proposed in this paper. In the first, the original signal with fault is decomposed into 6 intrinsic mode functions (IMFs) based on CEEMDAN. Second, the IMF contains the main characteristic information on the fault is selected based on the kurtosis index for next step. Finally, the fast kurtogram is applied on the chosen IMF to filter the signal and obtain the squared envelope spectrum. The results confirm that the proposed method can successfully extract the bearing fault signature form vibration signal.

14:15
Microwave Filter Optimization Combining Aggressive Space Mapping And Cauchy Methods
PRESENTER: Bouhafs Bouras

ABSTRACT. The aim of this research work is to demonstrate the efficiency of combining ASM and Cauchy methods through the optimal design of a fourth order microwave bandpass coaxial filter. The filter should give when tuned optimally the specified response respecting the specified parameters namely: center frequency, bandwidth and return loss. The tuning is the translation of these specifications into the actual physical dimensions of the filter, which is not a straightforward task. In contrast given a set of the physical dimensions the response and the parameters of the filter are easily calculated using simulation and extraction. The idea then is to start with some estimated dimension values, get the filter response of the filter by simulation, and apply Cauchy extraction technique to determine the actual non optimal parameters of the filter. The differences between these parameter values and the specified ones are used as error function by ASM method to correct for the physical dimensions. After simulation and extraction, parameters values that are closer to the specified ones are obtained. This two way procedure is repeated a certain number of iterations until a satisfactory match is achieved.

14:15
Low Cost RTL-SDR Based Positioning Solution Based on GSM Cell Id
PRESENTER: Reda Yagoub

ABSTRACT. The localization problem has several years of existence justified through a big number of research results on algorithms and implemented solutions. It comes to us, that developing and implementing embedded low cost universal localization and positioning solution, taking advantage of existing wireless communications infrastructures, are motivating, attracting and useful tasks. In the case of GSM mobile networks, several localization methods are proposed using the GSM network parameters such as Time Of Arrival (TOA), Angle of Arrival (AOA), etc. All proposed methods require a mobile phone or GSM module to listen within specific operator network to estimate the user location. The simple one is cell identity (Cell Id) positioning, which is not requiring an additional equipment or software upgrade. In this paper, we propose a new approach to improve the accuracy of Cell Id positioning method. The localization system suggested here, is based on RTL-SDR receiver and raspberry Pi, and do not require any subscriber identity module card (SIM), the cellular information for the all network operators in the region are extracted in the Raspberry Pi board and used to estimate the user’s position. The mean error of our proposed system is about 270 meters which is good if we take into consideration that we don’t have the exact position of the base stations (BTS).

14:15
Facial Kinship Verification using Tensor Discriminative Subspace based Color Components
PRESENTER: Ammar Chouchane

ABSTRACT. In this work, we propose the evaluation of color spaces of 2D facial images in order to determine the most discriminating color component for kinship facial verification in uncontrolled environments. Using multidimensional facial data modeling, our approach is based on higher order tensors. Multilinear sub-space projection is provided by a new method called Multilinear Side-Information Based Discriminant Analysis (MSIDA). Several inter-connected subspaces are obtained on different tensor modes, so that the subspaces are learned iteratively by unfolding the tensor in the different modes. Using only weak information, MSIDA projects the face tensor into a new subspace in which the discrimination is improved. The input of our system is one of the color spaces RGB, YcbCr and HSV,in which, each component is evaluated separately. The experimental evaluation in the Cornell KinFace database shows that the proposed approach outperforms significantly the most state-of-the-art approaches.

14:15
Design and FPGA real-time implementation of PWM and PPM modulation for Ultra Wide Band applications
PRESENTER: Hocine Hamil

ABSTRACT. Because of its high power efficiency, transmission efficiency and strong anti-jamming capability, PPM is widely used in spite the difficulty of synchronization between transmitter and receiver at the beginning of each symbol. The simplicity and the capacity of ultra wide Band initiate a lot of research activities from both academia and industry applications especially in wireless communication field. This paper discuss the FPGA implementation of PWM and PPM for Ultra Wide Band communications. The proposed design was simulated using Xilinx Vivado software tool and targeted on Nexys 4-DDR Board with Artix-7 FPGA device. Data input was generated using both DDS Compiler and Analog Discovery 2 with Pmod AD5 Analog to Digital Converter. Experimental results of the proposed design show that our approach is competitive to other related works and present a good compromise between low FPGA resources, low power consumption and high speed and can be widely used in many UWB applications.

14:15
Image Reconstruction based on Bayesian total variation and Hidden Markov Model
PRESENTER: Mansouri Boualem

ABSTRACT. Several methods of inversion in Fourier synthesis are based on interpolation of the data and fast inverse Fourier Transform (FT), but, the results obtained by such methods are not satisfactory when the data do not fill uniformly the Fourier domain which is the case in many applications in tomographic imaging because this inverse problem is known to be nonlinear and ill-posed. In this paper, we propose a regularization method based on the Bayesian estimation by introducing some prior information. In the first step, we assume that knowledge of the noise nature is considered as a prior information, for this and in order to improve the quality of image reconstruction, we use a votal variation prior whose model parameter is automatically determined using the Evidence Analysis within the Hierarchical Bayesian Paradigm. Secondly, we assume that the original image is composed of homogeneous regions, in this case we propose the Hidden Markov Modelling for image classification. This approach allows to have an algorithm that unifies restoration and reconstruction via classification in Fourier synthesis. This method is applied on synthetics and real images.

14:15
A Comparative Study of Hand Gestures Recognition Based on MobileNetV2 and ConvNet Models
PRESENTER: Khalil Bousbai

ABSTRACT. Hand gestures offer humans a natural way to interact with computers for a variety of applications. However, factors such as the complexity of hand gesture patterns, size differences, hand posture, and environmental lighting can affect the performance of hand gesture recognition algorithms. Recent advances in deep learning have greatly improved the performance of image recognition systems. In particular, the deep convolutional neural network has demonstrated superior performance in terms of image representation and classification, compared to conventional machine learning methods. This article proposes a two-technique comparison, a proposed deep-convolution neural network and transfer learning with the pre-trained model mobilnetv2 , for hand gestures recognition of American sign language. Both models are trained and tested using 1815 images segmented by colour and with black background, and its static hand gestures for five volunteers, which incorporate variations in features such as scale, lighting, and noise. The results show that, the proposed CNN model achieved a classification recognition accuracy of 98.9%. with a 2% improvement over the convolutional neural network model enriched using transfer learning techniques 97.06%.

14:15
Enhanced Performance of Substrate Integrated Waveguide Bandstop Filter Based on Metamaterials SCSRRs
PRESENTER: Kada Becharef

ABSTRACT. In this paper, two band-stop Substrate Integrated Waveguide (SIW) filters based on Square Complementary Split Ring Resonators (CSRRs) were presented for X-band applications by etching SCSR Resonators on the surface of the substrate integrated waveguide. Two topologies of the band stop filter are treated. The first toplogy concerns the use of one square ring CSRR are etched in the top plane of the SIW, the second includes three square rings CSRRs. All the structures are designed on a single substrate RT / Duroid 5880 with permittivity and tangential losses (tg(δ) = 0.0009). Numerical calculations using the Finite Element Method (MEF) based the High Frequency Structure Simulator (HFSS) software are presented to design this filter

14:15
APPLICATION OF METAMATERIALS FOR THE DESIGN OF ANTENNA
PRESENTER: Kada Becharef

ABSTRACT. Metamaterials have a very important role to play in new technologies, especially in the field of telecommunications. With the availability of the negative refractive index of metamaterials, the performance of microwave devices can be improved. These artificial devices have several advantages such as their integration with the planar environment and also their low cost of manufacture. In this article, we are interested in the design of volumic antennas in SIW using longitudinal slots and CSRR resonators. The results of the design of these antennas show an improvement in their performance in terms of adaptation and bandwidth.

14:15
ANSYS-HFSS Coupling Matrix Extraction with Immittance ‎Parameters for Microwave Bandpass Filter Tuning
PRESENTER: Rachid Bouhmidi

ABSTRACT. In this paper a new filter adjustment strategy based on AYSYS-HFSS simulations is proposed. The method relies on the introduction of additional internal ports into the filter design for the extraction of the multi-port admittance matrix (Y-Matrix). The main advantage is that the filter parameters, namely the resonant frequency of each resonator, the coupling coefficients and the external quality factors Qs are directly accessible via the multi-port Y-Matrix. Theextracted values will allow through a simple adjustment procedure to approach the specified data of the synthesis of the filter. An example of a folded bandpass coaxial filter with a four pole Chebyshev characteristicis chosen to demonstrate the validity of the proposed technique which can be adapted to other types of filters as well as to other characteristics.

14:15
Nonlinear Compensation in 32 Gbauds PM-16-QAM and PM-QPSK Coherent Optical Systems

ABSTRACT. The ultimate capacity of fiber-optic communication is limited by fiber nonlinearity. Advanced digital signal processing has the opportunity to relax this limitation. In this work, we investigate the compensation of fiber nonlinearities by digital backpropagation in both Polarization-division-multiplexed 16-level quadrature amplitude modulation system and polarization multiplexed 0.3 return-to-zero quadrature phase-shift keying system. The comparison is done at the same symbol rate. The systems performances are measured in term of bit error rate, optical signal noise ratio penalty, and the maximum reach. The polarization multiplexed 0.3 return-to-zero quadrature phase-shift keying system performs better than other system by increased sensitivity and the maximum transmission distance.

14:15
Sentiment analysis application on Twitter for E-reputation

ABSTRACT. The Internet has become the number one means of communication in the world using emails and social networks, it contains a large amount of data. Every Internet user can express himself freely and give his opinions and feelings, on any subject including companies. Those last ones are interested in improving their images, by analyzing the opinions presented in tweets or comments, and identifying weaknesses to correct them later.

This paper focuses on a new topic that is E-reputation or electronic reputation, where we will analyze the tweets on Twitter of a mobile operator in Algeria "Djezzy". The study is divided into two parts, the first part is the pre-treatment where we eliminate noise and keep only the necessary information, the second part is using Machine Learning techniques and algorithms to detect the company's reputation between Internet users, based on the three languages: French, English and Arabic.

14:15
Optimization of the gain flatness of EDFA in WDM PON system

ABSTRACT. the main purpose of this paper is to obtain the gain non-uniformity for each channel in order to equalize the amplitude gain in a WDM-PON system (Wavelength Division Multiplexing Passive Optical Network). The system is simulated to achieve gain flatness of EDFA (erbium doped fiber amplifier) through optimized fiber length and erbium concentration. The gains are flattened within 26.±0.7 dB from 1546 nm to 1558 nm band of wavelength with noise figure (NF) less than 5 dB , quality factor of 7.6 and bit error rate (BER) less than 10-9 for 8-channels simultaneous amplification.

14:15
Frequency Analysis Of Electromyogram Signals (EMGs)
PRESENTER: Fadia Meziani

ABSTRACT. The aim of this study is to analyse the electromyogram signals using frequency The aim of this study is to analyses EMG signals using frequency analysis. This analysis is based on frequency parameter such as median frequency and mean frequency by using the FFT and the non-linear information by using the bispectrum analysis , in order to provide a wide range of information’s related to the type of signal (normal and pathological)

14:15
Automatic Speech Recognition bay System ‎ MFCC-LPC-SVM for energy optimization ‎ in intelligent building
PRESENTER: Henni Sid Ahmed

ABSTRACT. the problem of smart buildings is their overconsumption of ‎energy; in this article we propose a solution to this problem. ‎This solution is to give optimum comfort to the people who ‎occupy the smart buildings, but sometimes these people are not ‎satisfied with this comfort, so we realized a system of automatic ‎speech recognition, to give the possibility for these people to ‎change the parameters of comfort, and to obtain the desired ‎comfort. This solution allows to optimize the smart building ‎energy, for example if a room is empty, then the person who ‎has been in this room must say off when leaving this room, in ‎this case the automatic recognition system of speech, instructs ‎the system that managed the energy to turn off the lamps and ‎lower the heater or air conditioner. This automatic speech ‎recognition system using the MFCC method (Mel frequency ‎cepstral coefficient) and a LPC method (linear prediction ‎coding) for the representation of the speech signal, and SVM ‎‎(support vector machine), for speech recognition. The ‎recognition will be of global type, the speech signal will be ‎parameterized using the MFCC coefficients and LPC ‎coefficients to form the input vector of the recognition system, ‎the SVM will be used for the learning phase and recognition ‎phase. An experiment will be implemented using an input ‎corpus, the vocabulary used in this corpus consists of 10 words ‎in English language. The results obtained will be analyzed and ‎compared with other similar works that use other techniques ‎of automatic speech recognition.‎

14:15
Awatermarking image algorithmusing Edge insertion and Symmetric Partition Algorithm(SPA)

ABSTRACT. This article contains a new algorithm that can be summarized into a symmetrical division of images and using a linear function to embed the watermark image in a spatial domain. The exponential function are used to adjust the value of the pixel during insertion to achieve robustness against attacks. To obtain more satisfactory results compare with other methods in the spatial domain. The proposed algorithm that was created defends safety, efficiency and capacity as well as simplicity and rapidity. This mathematical model is experimented against most unintentional attacks and demonstrated us acceptable results compared to other existing research.

14:15
A New Design of a THz Ʌ-Helical Antenna based on ‎MEMS Technology using SNLPs

ABSTRACT. Micro-Electromechanical Systems (MEMS) technology has been ‎efficiently applied in modeling a large class of terahertz (THz) ‎antennas due to the fast progress of Computer Aided Design ‎‎(CAD) tools which meet the very specific requirement of such ‎compact antennas thanks to advanced electromagnetic ‎optimization techniques. This paper presents a novel design of a ‎MEMS helix antenna developed using complex three-‎dimensional High Frequency Structure Simulator- (3D HFSS) ‎based Sequential Non Linear Programming (SNLP) algorithms ‎to achieve a high geometrical accuracy and provide a high ‎performance for the selective band of frequencies by rectifying ‎various tunable parameters embedded in silicon platform ‎including both helix parameters and feeding line characteristics. ‎The helix antenna using MEMS technology shows excellent ‎electromagnetic response demonstrated by very low reflection ‎coefficients of less than -10 to -45dB, excellent voltage standing ‎wave ratios of less than 1.20 to 1.00, and very good directivity ‎for a wide range of frequencies extending from 800 GHz to 1.05 ‎THz.‎

14:15
OTDM Transmission Multiplexing and Demultiplexing for Photonic Packet Switched (PPS) Networks

ABSTRACT. This work is part of optical telecommunications field. It is to study the Packet Photonic Swiching (PPS) OTDM transmission technique. The PPS-OTDM is a system for generating pulse trains and at high repetition rates beyond the limit of the electronic bandwidth by a packet time interleaving process. However, the demultiplexing system requires an « AND » gate. The amplitude modulator « AM » for its property of multiplying the signals is best suited for this demultiplexing system. Knowing that the OTDM requires a source of emission and a detection technique adapted to the reception.The simulation of the complete transmission chain was done under the software Optisystem version 9.0 for an 8-bit packet system with a bit rate of 2.5 Gb / s and a wavelength of 1550nm.

14:15
Multimodal Medical Images Registration Using Biogeography-Based Optimization Algorithm
PRESENTER: Charif Fella

ABSTRACT. Image registration plays an important role in medical analysis. Nowadays, it constitutes a very active research field due to the great improvement of the medical imaging technology. Metaheuristics based algorithms such as Genetic Algorithm (GA) , Particle Swarm Optimization (PSO) , Artificial Bee Colony (ABC) and Differential Evolution (DE) have been widely applied to medical image registration. Due to the success of the Biogeography-based Optimization (BBO) in the global optimization problems, we present in this paper an improved method for medical image registration using the Positron Emission Tomography (PET) and Magnetic Resonance (MR) modalities. Experimental results demonstrate that the proposed approach achieves high-precision and robust registration compared to the PSO algorithm.

14:15
Deep convolutional neural networks for detection and classification of tumors in mammograms
PRESENTER: Kadda Djebbar

ABSTRACT. Abstract—Automatic detection and classification of the masses in mammograms are still a big challenge and play a crucial role to assist radiologists for accurate diagnosis. In this paper, we propose a novel Computer-Aided Diagnosis (CAD).System based on one of the regional deep learning techniques, a ROI- based Convolutional Neural Network (CNN) which is called You Only Look Once (YOLO), in this work we used YOLO version three (YOLOv3). YOLO based CAD system can handle detection and classification simultaneously in one framework. It’s a little bigger than last time but more accurate. The proposed CAD system contains four steps : preprocessing of mammograms, feature extraction utilizing deep convolutional networks, mass detection with confidence, and finally mass classification using fully connected neural networks (FC-NNs). Index Terms—Breast Cancer; Mass Detection and Classification; Computer Aided Diagnosis; Deep Learning; YOLO

15:15-15:30Coffee Break
15:30-18:00 Session 7A: Signal processing
15:30
Gradient Based Neural Network for Superfast Capon Spectral Estimation

ABSTRACT. This paper describes the use of a novel gradient based recurrent neural network to perform Capon spectral estimation. Nowadays, in the fastest algorithm proposed by Marple et al., the computational burden still remains significant in the calculation of the autoregressive (AR) Parameters. In this paper we propose to use a gradient based neural network to compute the AR parameters by solving the Yule-Walker equations. Furthermore, to reduce the complexity of the neural network architecture, the weights matrix-inputs vector product is performed efficiently using the fast Fourier transform. Simulation results show that proposed neural network and its simplified architecture lead to the same results as the original method which prove the correctness of the proposed scheme.

15:50
Statistical Analysis and New Modeling of Real Clutter Signal in FM Radio-based Passive Radars
PRESENTER: Linda Bendjama

ABSTRACT. The major problematic in the design of a competitive detector is the construction of a statistical model that fits real data. In this paper, we present a new theoretical model that describes real data from an FM (Frequency Modulation) radio-based passive bistatic radar. In doing this, we aim to estimate the model parameters using a conventional technique and two optimization numerical methods; namely, the trust-region-reflective approach for non-linear minimization subject to bounds and the direct search Method based on the Nelder-Mead (N-M) algorithm..

16:10
Effect of Fractional Order Moments on Parameter Estimation of K-Clutter plus Noise

ABSTRACT. Parameter estimation of radar clutter is considered as a critical task for the development of target detectors. This work covers the estimation of the shape parameter of K-clutter plus noise using a modified scheme of fractional order moments method (FOME). If the fractional order is constant, the FOME has a closed form and undesirable estimation errors are produced in some cases. In order to achieve better estimation performance, the fractional order and the shape parameter should be optimized together. To this effect, a numerical formula of the corresponding fitness function is given and unconstrained nonlinear optimization method based on the Nelder-Mead simplex algorithm is used to compute the unknown parameters. Via simulated K-clutter plus noise data, the effect of the fractional order on the estimation accuracy is studied firstly. Then, to show the efficiency of the proposed estimator, comparisons with existing HOME, FOME using fixed fractional order moment and [zlog(z)] methods are conducted.

16:30
Experimental Design and Analysis of Sound Event Detection Systems: Case Studies
PRESENTER: Hadjer Ykhlef

ABSTRACT. Sound Event Detection (SED) systems have attracted a widespread attention from the Machine Learning community due to their potential applications. As in many sound analysis scenarios, it is of paramount importance to assess and compare the performances of such systems trained from sound data. The main goal of system evaluation and comparison is to derive conclusions unaffected by chance and are therefore significant. Both the design of experiments and the analysis of results through statistical tests should be conducted properly in order to insure significance. In this paper, we first discuss the principles of machine learning experiments in the context of SED. Then, we show the proper methodologies through case studies. To this end, we have examined four classification models (Support Vector Machine, Convolutional Neural Network, Adaboost and Random Forest) trained using Mel Frequency Cepstral Coefficients (MFCCs). Most importantly, we have supported our analysis and discussions with numerous statistical tests. Without the use of a proper data augmentation approach, the experimental results indicate the superiority of the ensemble-based classifiers (Random Forest and Adaboost), with an overall detection accuracy of 83%. Furthermore, adding the first and second derivative of MFCCs significantly improves the performance of SVM-based systems.

16:50
New Measurement Method for Respiration And Heartbeat Estimation Using IR-UWB Radar
PRESENTER: Ibrahim Kakouche

ABSTRACT. In this work, the feasibility of using Delay And Sum (DAS) method in frequency domain and Variational Mode Decomposition algorithm (VMD) for respiration rate And heartbeat estimation using Impulse Radio Ultra-Wideband (IR-UWB) radar is demonstrated. The proposed method analyses the velocity of the chest displacement which is associated with breathing and heartbeat of human body. It based on solving two problems : velocity estimation and signal separation. The received matrix data is divided into a set of Coherent Processing Integration (CPI) times in slow time direction. To extract the vital signs signals,vDAS method in frequency domain is adopted to IR-UWB radar to estimate the velocity of chest movement for each CPI. VMD algorithm is then applied to separate the Heartbeat signal from the respiratory signal. Experimental results proved the feasibility and the reliability of the proposed approach.

17:10
Automatic identification and analysis of epileptic seizure
PRESENTER: Manel Yakoubi

ABSTRACT. Epilepsy is a common neurological disease that is characterized by the recurrence of seizures. Electroencephalogram (EEG) signals are widely used to diagnose epilepsy. This paper presents an automated diagnostic method for epileptic detection using gaussian mixture model based statistical time domain features. The statistical time domain namely root mean square (RMS), Sekwens (Sk), kurtosis (Ku), crest factor (CF) and clearance indicator (CLI) parameters are extracted. These features are used to train gaussian mixture models (GMM) for classification of EEG signals into three classes normal, pre-ictal and ictal. The GMM classifier is able to discriminate the three types of EEG signals and achieved better results compared to other existing techniques with an average accuracy of 98.75%. The proposed work can be used for automatic diagnosis of seizure onsets which would help the patients to take appropriate precautions.

17:30
EEG Signal Classification for Epileptic Seizures Detection Based on AR Models
PRESENTER: Belhadj Sabrina

ABSTRACT. In this study a novel approach for automatic epileptic seizure detection using EEG signals is introduced. Firstly, the autoregressive model parameters have been calculated using Yule-Walker and the Burg’s methods. Then, a fusion vectors have been constructed using the score level fusion of the coefficients obtained using Yule-Walker and burg’s methods, three cases of weights have been investigated. The Akaik Information Criterion (AIC) has been employed to select the model order. Finally, the fusion vectors have been used as feature to classify the EEG signals into different classes using SVM classifier. Experimental tests have been yielded in order to estimate the performance of the presented framework. The obtained results with an average accuracy for the first, second and the third cases are between 89.50% and 98.50%. The results are very interesting and have clearly demonstrated the value of the proposed framework in classifying the different groups of the EEG signals. The proposed strategy can be successfully used to discriminate between healthy and epileptic subject.

15:30-18:00 Session 7B: Applications
15:30
Impact of Metamaterials DGS in PIFA Antennas for IoT Terminals Design

ABSTRACT. In this paper, the impact of the Defected Ground Structure (DGS) in PIFA antennas for IoT terminals is evaluated. Numerical analysis of antenna performance using finite element methods in term of gain and matching are discussed for Fishnet based PIFA antenna. The proposed PIFA antenna configuration is based on Fishnet Metamaterials etched in the ground layer using Defected Ground Structure (DGS)technique. The simulation results of the proposed Fishnet based PIFA antenna show dual band frequency at 3.9 GHz and 5 GHz with an enhancement in return loss, a slightly higher total efficiency of 100% and a gain enhancement of 3.11dB compare to the classical PIFA antenna gain.

15:50
PAPR Reduction in MIMO (2x2)-FBMC-OQAM Systems Using Attenuating QAM Symbols
PRESENTER: Hocine Merah

ABSTRACT. In this paper, a scheme named Attenuating Quadrature Amplitude Modulation Symbols (AQAMS) with Multi-input Multi-output (MIMO) system is proposed to reduce the Peak-to-Average Power Ratio (PAPR) of the Filter Bank Multi-Carrier (FBMC) system adopting Offset Quadrature Amplitude Modulation (FBMC-OQAM) structures. A remarkable strategy has been followed mainly relying on multiplying every Quadrature Amplitude Modulation (QAM) by real decreasing magnitudes, then the transmitted signal becomes closer to a constant cover i.e. reduced in PAPR features. Simulation findings following the AQAMS technique illustrate concrete progress regarding PAPR reduction.

16:10
Statistical features extraction in wavelet domain for texture classification
PRESENTER: Soraya Zehani

ABSTRACT. This paper presents a new approach for texture classification generalizing a well-known statistical features combining the fractal analysis by means of fractal dimension (FD) with the selection first and second order statistics features in the spatial and wavelet domain. The objective of our paper is to propose the features extarction using statistical parameters in the saptial domain and in wavelet domain with different wavelets, with and without preprocessing stage for the texture classification using pattern recognition neurals network and studying the effect of the preprocessing and the wavelets in the classification rate. The extracted features are used as the input of the ANN classifier. The performance of the proposed methods evaluated by using two classes of Brodatz database textures and has been compared. The proposed methods are evaluated using confusion matrix and ROC (receiver operating characteristics) curves.

16:30
SAX2SEX: Gender Classification on 3D Faces using Symbolic Aggregate ApproXimation
PRESENTER: Samia Bentaieb

ABSTRACT. Gender classification is a demographic attribute that found an increasing amount of applications particularly in human-computer interaction, security access control and biometrics. The purpose of this paper is to investigate the feasibility of using time series for gender classification. After transforming input depth images of faces into one-dimensional vector using Peano-Hilbert space filling curve, a dimensionality reduction step is performed to convert this vector to a sequence of alphabets using Symbolic Aggregate approXimation (SAX). With FRGCv2 dataset, standard 10-fold cross-validation experiments are performed using quadratic kernel for SVM classifier. The results have shown that the proposed approach can reach a correct classification rate higher than 96%.

16:50
Robust and reliable PPG and ECG integrated biosensor
PRESENTER: Merzak Ferroukhi

ABSTRACT. Noncommunicable diseases (), also known as chronic diseases (cancers, respiratory and cardiovascular diseases, diabetes…), tend to be of long duration and are the result of a combination of genetic, physiological, environmental and behaviors factors. According to the World Health Organization, cardiovascular diseases account for most deaths or 17.9 million people annually. It is well-known that high blood pressure (hypertension) is a frequent and major cardiovascular risk factor involved in the development of stroke, ischemic heart disease, heart failure and dementia. In addition, the estimation of blood pressure can be achieved by determining an important physiological parameter called  pulse transit time (PTT), which is the time required for a pulse wave to move between two different arterial points. In this work, we propose three main challenges in embedded systems design: the challenge to build a simple but reliable system and that to build a robust and low-cost system that simultaneously acquires electrocardiogram () and photoplethysmography () signals to measure . This system is very useful in outpatient medicine. The ultimate objective to be achieved is to make our prototype smart by using machine learning techniques, and miniaturized as part of the Internet of Things (), which opens up new possibilities in medicine that can give additional insight into symptoms and trends, enable remote care and, in general, give patients more control over their lives and their treatment.

17:10
Real-time Aging Friendly fall detection system
PRESENTER: Hafida Saidi

ABSTRACT. The fall is a crucial issue for elderly people. With the strong growth of information and communication technologies and the development of light and low-cost wearable technology, elderly fall detection has gained much attention. These technologies can help the elderly get timely assistance to reduce further injury. In this paper, we propose to develop a real-time aging friendly fall detection system based on angular velocity and acceleration values produced by the accelerometer and gyroscope sensors of the mobile device. We describe an effective and simple threshold-based solution to implement a real-time method to monitor the elderly and detect falls using a mobile phone. In case of a fall, our system will transmit an alert with location information to the contacts list via a notification. Thus, medical attention can be provided with minimal delay. The system was tested by volunteers and achieved high sensitivity, specificity and accuracy.

17:30
Detecting falls with recurrent autoencoders and body acceleration data

ABSTRACT. According to the world population statistics, the elderly proportion is continuously growing. With this growth, the concerns of this population increase. Falls are a common problem within the elderly population causing considerable health and financial impacts. In order to mitigate such impacts, automatic fall detection systems are more and more employed by the elderly. In this work, we propose a fall detection system based on the body acceleration data, where we investigate the use of recurrent-autoencoders to distinguish between normal activities of daily living and falls. We study different training strategies combined with either the use of the autoencoder as feature extractor or employing its reconstruction error for decision. Experiments on the UniMiB SHARE fall data set provides interesting insights about the designed system.