Résumé: Prolonging the lifetime of wireless sensor networks (WSN) is the biggest challenging issue. In wireless multimedia sensor networks (WMSNs), sensor nodes have limited energy resource and to increase the lifetime of the network, it is necessary to design an effective fast algorithm that aims at reducing the consumed power. This paper proposes an energy-efficient discrete cosine transform (DCT) approximation requiring only 12 additions. Associated with a JPEG compression chain, this DCT approximation ensures a very good rate-distortion compromise, but above all, a very low computational complexity and significant compatibility with the exact DCT. Simulation results clearly show that the proposed fast transform algorithm achieves a better trade-off between image quality, computational complexity and energy consumption compared to any existing pruned DCT approximations. Furthermore, it is very suitable for the resource constrained wireless visual sensor networks (WVSNs) requiring low bitrates.

Résumé: The recent technological advances in surveillance, forensic and biometric systems to deter or even reduce the increasing number of crimes and prevent them is still questionable. The use of gait biometrics has attracted unprecedented interest due to its capability to work with low‐resolution footage recorded from a distance. In contrast to mainstream research on gait biometrics which uses holistic silhouette features, the authors investigate the use of the bottom dynamic section within the human body to derive the most discriminative features for gait recognition. A new descriptor based on 7 Hu's moments is proposed describing the inner lower limb regions between the limbs being extracted only from landmark frames within one gait cycle. In order to assess the discriminatory potency of gait features from the lower regions for people identification, a number of experiments are conducted on the CASIA‐B gait database to investigate the recognition rates using the KNN classifier and deep learning. The comparative analysis is performed against well‐established research studies which were tested on the CASIA‐B data set. The obtained results confirm the consistency of features extracted from the lower regions for gait recognition even under the impact of various factors.

Résumé: Vehicular communication has become a reality guided by various applications. Among those, high video quality delivery with low latency constraints required by real-time applications constitutes a very challenging task. By dint of its never-before-achieved compression level, the new High-Efficiency Video Coding (HEVC) is very promising for real-time video streaming through Vehicular Ad-hoc Networks (VANET). However, these networks have variable channel quality and limited bandwidth. Therefore, ensuring satisfactory video quality on such networks is a major challenge. In this work, a low complexity cross-layer mechanism is proposed to improve end-to-end performances of HEVC video streaming in VANET under low delay constraints. The idea is to assign to each packet of the transmitted video the most appropriate Access Category (AC) queue on the Medium Access Control (MAC) layer, considering the temporal prediction structure of the video encoding process, the importance of the frame and the state of the network traffic load. Simulation results demonstrate that for different targeted low-delay video communication scenarios, the proposed mechanism offers significant improvements regarding video quality at the reception and end-to-end delay compared to the Enhanced Distributed Channel Access (EDCA) adopted in the 802.11p. Both Quality of Service (QoS) and Quality of Experience (QoE) evaluations have been also carried out to validate the proposed approach.

Résumé: Reducing the algorithmic complexity of image compression techniques is a major challenge in wireless image sensor networks (WISNs). Many image compression standards, such as JPEG and JPEG2000, are unsuitable for implementation in WISNs because of their high energy consumption. In this paper, a solution to this problem is proposed. It consists of a region-of-interest (ROI) based image compression using the discrete Tchebichef transform (DTT).The main idea is about compressing only the ROI instead of the whole image. The DTT is used as an alternative to the discrete cosine transform (DCT) due to its low complexity and good energy compaction. Simulation results have shown that the proposed method reduces the number of arithmetic operations, the processing/transmission energy consumption and the amount of transmitted data. The savings obtained generally exceed 50%. Furthermore, it has a competitive compression efficiency compared with the state-of-the-art image compression techniques.

Résumé: There is nowadays a growing demand in vehicular communications for real-time applications requiring video assistance. The new state-of-the-art high-efficiency video coding (HEVC) standard is very promising for real-time video streaming. It offers high coding efficiency, as well as dedicated low delay coding structures. Among these, the all intra (AI) coding structure guarantees minimal coding time at the expense of higher video bitrates, which therefore penalizes transmission performances. In this work, we propose an original cross-layer system in order to enhance received video quality in vehicular communications. The system is low complex and relies on a multiple description coding (MDC) approach. It is based on an adaptive mapping mechanism applied at the IEEE 802.11p standard medium access control (MAC) layer. Simulation results in a realistic vehicular environment demonstrate that for low delay video communications, the proposed method provides significant video quality improvements on the receiver side.

Résumé: Orthogonal Frequency Division Multiplexing (OFDM) is of great interest for the development of the fifth-generation technology. It is the cornerstone of Multiple-input multiple-output (MIMO) systems. Even though inter carrier interference (ICI) and inter symbol interference (ISI) have been processed for the fourth-generation standards, they still present a huge problem for the fifth-generation standards. This paper explores the tradeoff between the length of the cyclic prefix and the performances of the OFDM system. It also studies the effect of carrier frequency offset (CFO) on OFDM systems. A blind frequency offset estimator that uses the correlations between the remodulated sequence in the receiver side and the conventional received symbol is presented and a closed form solution is derived. The proposed estimator is derived under short interval when the correlation is high, so it has low computational complexity. Lin and Beek’s estimators are used for comparison. Simulations demonstrate the effectiveness of the proposed estimator under Rayleigh fading channel.

Résumé: To benefit the properties of both Low-Density Parity-Check (LDPC) and Turbo Convolutional Codes (TCC), we propose a practical concatenated Gallager/Convolutional code in a turbo coding way. The modified code creates a balance between the advantages and the disadvantages of LDPC and TCC in terms of the overall complexity and latency. This will be done through two different component decoders; and convolutional code of the same rate without interleaver. Since the two decoders are different in nature, they exchange extrinsic information that will be easily adapted to each other. The study of computation complexity and decoding performance over an AWGN channel indicates that such approach leads to excellent performance because of several factors. The proposed approach achieves a trade-off between waterfall and error floor regions. It reduces complexity decoding compared to TCC and 3D-TCC. It provides a better coding gain over LDPC and PGCG (Parallel Concatenated Gallager Codes). These features will ensure optimal outcomes and cost-performance ratio, and thus, this trend can be the best choice for today's communication systems.

Résumé: Spectral amplitude coding for optical codedivision multiple access (SAC-OCDMA) networks hasreceived much attention over the last two decades. Thisarticle aims to explore the impact of encoder change ondifferent types of optical filters, such as the Gaussian opticalfilter and the Bessel optical filter, for high data rates and togive an overview on importance of choosing the optimaltype of optical filter according to the frequency rangeselected by the user is 25 and 50 GHz. SAC-OCDMA trans-mitter utilizes Wavelength Division Multiplexing multi-plexer (WDM MUX) as an encoder, to generate a codehaving low cross-correlation called Random Diagonalcode, and spectral direct detection as a detection technique.The change of optical filter, in WDM MUX, directly affectsthe performance of the system. The results show that thesystem for 50 GHz, with a WDM MUX, using a Gaussianoptical filter has better performance compared to the opticalBessel filter and can reach a bit error rate (BER) of 10−25.SAC-OCDMA system, using a WDM MUX based on Besselfilters with a bit rate of 300Mb/s, achieves a BER of 10−28which leads us to recommend it for second norm 25 GHz.Moreover, the power received increases by 4 dBm every 20Km with the increase in the length of the fibre for both filtersBessel and Gaussian. Our work focuses on the two 25 and50 GHz bands, after a study on the impact of the change ofthe bandwidth and the order of the different optical filtersused according to the BER applied to the different networksof access, such as local area network (LAN)

Résumé: Recently, video streaming has attracted much attention and interest due to its capability to process and transmit large data. We propose a quality of experience (QoE) model relying on high efficiency video coding (HEVC) encoder adaptation scheme, in turn based on the multiple description coding (MDC) for video streaming. The main contributions of the paper are (1) a performance evaluation of the new and emerging video coding standard HEVC/H.265, which is based on the variation of quantization parameter (QP) values depending on different video contents to deduce their influence on the sequence to be transmitted, (2) QoE support multimedia applications in wireless networks are investigated, so we inspect the packet loss impact on the QoE of transmitted video sequences, (3) HEVC encoder parameter adaptation scheme based on MDC is modeled with the encoder parameter and objective QoE model. A comparative study revealed that the proposed MDC approach is effective for improving the transmission with a peak signal-to-noise ratio (PSNR) gain of about 2 to 3 dB. Results show that a good choice of QP value can compensate for transmission channel effects and improve received video quality, although HEVC/H.265 is also sensitive to packet loss. The obtained results show the efficiency of our proposed method in terms of PSNR and mean-opinion-score.

Résumé: High-efficiency video coding (HEVC) is based on integer discrete cosine transforms (DCTs) of size 4 × 4, 8 × 8, 16 × 16 and 32 × 32 whose elements are coded on 8 bits. However, the algorithm requires that the output length at each processing stage should never exceed 16 bits. The conventional solution is the truncation of the least significant bits which leads to erroneous results and a waste of resources. In this Letter, different DCT kernels with reduced elements length are proposed. They are evaluated for compression efficiency and hardware implementation competence. An implementation on the Xilinx FPGA virtex-6 circuit has given a maximal operating frequency increase of 4.81 and 93.41% for the DCT-II 4 × 4 and the discrete sine transform, while reducing the energy consumption by 10.64 and 30.77% at 100 MHz, respectively. Using the HM 16.3 HEVC model and video sequences of different resolutions, the results show a quality degradation of 0.01 dB for a bit rate increase of 0.19% compared to the reference cores.

Résumé: In this paper, we propose an electrocardiogram (ECG) signal compression algorithm that is based on wavelet and a new modified set partitioning in hierarchical trees (SPIHT) algorithm. The proposed algorithm contains a preprocessing of the approximation subband before the coding step by mean removing. Three other modifications are also introduced to the SPIHT algorithm. The first one is a new initialization of the two lists of insignificant points (LIP) and insignificant sets (LIS), while the second is concerning the position of inserting new entries of type A at the LIS, and in the last one, the redundancy in checking type B entries in the original method was found and avoided. The new proposed coding algorithm is applied to ECG signal compression and the obtained numerical results on the MIT-BIH database show the efficient performances of the proposed SPIHT algorithm over the original method and other existing methods.

Résumé: In this paper we propose an iterative parallel decision feedback (P-DF) receivers associated with parallel interference cancellation (PIC) for multicarrier code division multiple access (MC-CDMA) systems in a Rayleigh fading channel (cost 207). First the most widely detection techniques, minimum mean-squared error MMSE, Maximum Likelihood ML and PIC were investigated in order to compare their performances in terms of Bit Error Rate (BER) with parallel feedback detection P-DFD. A MMSE DF detector that employs parallel decision-feedback (MMSE-P-DFD) is considered and shows almost the same BER performance with MMSE and ML, which present a better result than the other techniques. In a second time, an iterative proposed method based on the multi-stage techniques P-DFD (parallel DFD with two stages) and PIC was exploited to improve the performance of the system.

Résumé: In this paper, a new real-time approach for audio recognition of waterbird species in noisy environments, based on a Texas Instruments DSP, i.e. TMS320C6713 is proposed. For noise estimation in noisy water bird's sound, a tonal region detector (TRD) using a sigmoid function is introduced. This method offers flexibility since the slope and the mean of the sigmoid function can be adapted autonomously for a better trade-off between noise overvaluation and undervaluation. Then, the features Mel Frequency Cepstral Coefficients post processed by Spectral Subtraction (MFCC-SS) were extracted for classification using Support Vector Machine classifier. A development of the Simulink analysis models of classic MFCC and MFCC-SS is described. The audio recognition system is implemented in real time by loading the created models in DSP board, after being converted to target C code using Code Composer Studio. Experimental results demonstrate that the proposed TRD-MFCC-SS feature is highly effective and performs satisfactorily compared to conventional MFCC feature, especially in complex environment.

Résumé: After the successful development of the discrete cosine transform (DCT) especially in audio, image and video coding, many state-of-the-arts DCT-based fast algorithms have been developed. However, these algorithms are floating-point and are implemented with high algorithmic complexity and memory intensive. Therefore, they consume large system resources. In particular, those requirements pose a serious limitation on multimedia applications in resource-constrained systems, such as wireless multimedia sensor networks (WMSNs). In this paper we investigate an alternative to DCT transform which known as discrete Tchebichef transform (DTT). Despite its good energy compaction property, low algorithmic complexity and low memory requirements, the DTT is potentially unexploited in the literature. To further reduce its complexity and memory requirements, in this paper we propose a pruned version of DTT (PDTT). Simulation results show that P-DTT requires a reduced number of arithmetic operations, energy consumption and memory. And it has at the same time competitive compression efficiency compared with Loeffler-DCT and exact DTT. The proposed P-DTT transform is a viable option for image compression and/or communication over WMSNs.

Résumé: Image representation in separable orthogonal basis cannot take advantage of geometrical regularity contained in basic images. When, explored efficiently geometrical regularity improves image compression. In this paper, we propose to experiment and compare an adaptive multiscale geometric decomposition for Synthetic Aperture Radar (SAR) image compression, called multiscale Bandelet transform, and a non adaptive multiscale geometric representation called Ridgelet transform. The second generation of Bandelet transform adopted in this work, is constructed in discrete domain with bandeletization of warped wavelet transform along the optimal direction of geometric flow that minimizes the Lagrangian. We discuss the criteria and results to assess SAR image compression performances using wavelet, Bandelet, and Ridgelet transforms. Our experiments revealed that during the compression phase, the speckle noise is removed from the SAR images inducing further improvements of the coding efficiency. In order, to evaluate the robustness of Bandelet transform, we have proposed a progressive compression scheme based on the second generation of Bandelet transform combined to SPIHT encoder, which is generally integrated with the wavelet transform. Keywords: SAR image compression; bandelet transform; geometrical flow; ridgelet transform; wavelet; SPIHT encoder

Résumé: This paper presents a performance evaluation of the high efficiency video coding (HEVC) based exclusively on the discrete cosine transform DCT-II and compares it to the reference algorithm which uses with the DCT-II, a discrete sine transform (DST-VI) for 4×4 intra-predicted luminance blocks. Then, a modified version of DST-VI called “MDST-VI” is proposed and evaluated. Using the HEVC reference model HM-16.3 and the main profile, the test results show that the MDST-VI presents an interesting compromise between the DST-VI and the DCT-II. The former leads to the best rate-distortion results, the latter yields a quality mean decrease of 0.02 dB and a mean bit-rate increase of 0.47 %, while the MDST_VI reduces the previous figures to the tenth. In a second part, a FPGA implementation performance analysis of the three transforms indicates that the DCT-II and the MDST-VI outperform the DST-VI with a 55.27 and 41.35 % area reduction, 47.73 and 10.61 % energy saving, and 85.70 and 10.55 % throughput gain, respectively. The MDST-VI could well fulfill the requirements of present and next-generation mobile communication where high-throughput and low-powered hardware accelerators are highly desirable while ensuring the best quality and the lowest bit-rate. Keywords Discrete cosine transform (DCT)Discrete sine transform (DST)Field programmable gate arrays (FPGA)High efficiency video coding (HEVC)

Résumé: The key solution to study birds in their natural habitat is the continuous survey using wireless sensors networks (WSN). The final objective of this study is to conceive a system for monitoring threatened bird species using audio sensor nodes. The principal feature for their recognition is their sound. The main limitations encountered with this process are environmental noise and energy consumption in sensor nodes. Over the years, a variety of birdsong classification methods has been introduced, but very few have focused to find an adequate one for WSN. In this paper, a tonal region detector (TRD) using sigmoid function is proposed. This approach for noise power estimation offers flexibility, since the slope and the mean of the sigmoid function can be adapted autonomously for a better trade-off between noise overvaluation and undervaluation. Once the tonal regions in the noisy bird sound are detected, the features gammatone teager energy cepstral coefficients (GTECC) post-processed by quantile-based cepstral normalization were extracted from the above signals for classification using deep neural network classifier. Experimental results for the identification of 36 bird species from Tonga lake (northeast of Algeria) demonstrate that the proposed TRD–GTECC feature is highly effective and performs satisfactorily compared to popular front-ends considered in this study. Moreover, recognition performance, noise immunity and energy consumption are considerably improved after tonal region detection, indicating that it is a very suitable approach for the acoustic bird recognition in complex environments with wireless sensor nodes. Keywords Birdsong recognitionGTECCQCNNoise power estimationDeep neural networkWSN

Résumé: Energy consumption is a critical problem affecting the lifetime of wireless image sensor networks (WISNs). In such systems, images are usually compressed using JPEG standard to save energy during transmission. And since DCT transformation is the most computationally intensive part in the JPEG technique, several approximation techniques have been proposed to further decrease the energy consumption. In this paper, we propose a low-complexity DCT approximation method which is based on the combination of the rounded DCT with a pruned approach. Experimental comparison with recently proposed schemes, using Atmel Atmega128L platform, shows that our method requires less arithmetic operations, and hence less processing time and/or the energy consumption while providing better performance in terms of PSNR metric. Keywords: DCT Approximation; computational complexity; image compression; wireless sensor networks

Résumé: Apart from the efficient compression, reducing the complexity of the view random access is one of the most important requirements that should be considered in multiview video coding. In order to obtain an efficient compression, both temporal and inter-view correlations are exploited in the multiview video coding schemes, introducing higher complexity in the temporal and view random access. We propose an inter-view prediction structure that aims to lower the cost of randomly accessing any picture at any position and instant, with respect to the multiview reference model JMVM and other recent relevant works. The proposed scheme is mainly based on the use of two base views (I-views) in the structure with selected positions instead of a single reference view as in the standard structures. This will, therefore, provide a direct inter-view prediction for all the remaining views and will ensure a low-delay view random access ability while maintaining a very competitive bit-rate performance with a similar video quality measured in peak signal-to-noise ratio. In addition to a new evaluation method of the random access ability, the obtained results show a significant improvement in the view random accessibility with respect to other reported works.

Résumé: Multi-view video has attracted considerable interest due to its wide use in a growing market including 3D television, free viewpoint video and intelligent surveillance. Besides the efficiency of compression methods, a vital requirement for multi-view video coding procedures is the view random access which is described as the capability to navigate quickly to any arbitrary view at any given time. In this paper, a new prediction structure based on the increase of the hierarchical level of B-views is proposed. This approach involves reducing the number of images needed for the prediction of B pictures for specific B-views. To show the effectiveness of the proposed prediction structure, a new metric for evaluating the view random access is described. In contrast to the metric proposed by the JVT group which is limited to consider only the image having the maximum number of frames needed for decoding, the key basis of the proposed metric is to consider evaluating all the images contained within a Group of Group Of Pictures. Experimental results have shown that compared with the IBP prediction structure of the reference model JMVM, the proposed algorithm improves the view random access by up to 33.5% with significant improvement in terms of bit rate.

Résumé: The increasing demand to develop a palmprint biometric system with a low-error rate has prompted scientists to use multispectral imaging to overcome the limits of the techniques that act in visible light. In order to improve the accuracy of multispectral palmprint recognition, we explore two level fusions: pixel and the feature level fusion approaches. The former is based on a maximum selection rule, which combines discriminating information from different spectral bands of discrete wavelet transform of multispectral images. The latter operates the fusion of features extracted from subimages. We propose to use both approaches for statistical and energy distribution analysis of the finite ridgelet transform coefficients, for the sake of their simplicity and low-computational complexity. Once the feature vectors are obtained, we perform a robust classification to identify/verify individuals with both approaches. The effectiveness of the proposed methods is evaluated on several classifiers for binary and multiclass cases. The experimental results conducted on Chinese Academy of Sciences Institute of Automation and Hong Kong Polytechnic University databases show that the proposed approaches ensure, respectively, an accuracy rate of 100% and 99.79%. A comparative study has revealed that our approach outperforms or at least equals the performances of the state-of-the-art multispectral palmprint recognition methods.

Résumé: Studies of dysarthric speech rhythm have explored the possibility of distinguishing healthy speakers from dysarthric ones. These studies also allowed the detection of different types of dysarthria. The present paper aims at assessing the ability of rhythm metrics to perceive dysarthric severity levels. The study reports on the results of a statistical acoustic investigation using various rhythmic metrics. Among these rhythm features, we propose a new rhythm metric based on an approximation of the speakers’ rate of articulation. The investigation was carried out on the speech data of US dysarthric patients recorded on the Nemours corpus. The rhythm features are based on two types of segmentation: vocalic/consonantal and voiced/unvoiced interval durations. Results of different classification experiments show that the rhythm-based measures can be used effectively to characterise the dysarthric severity by classifying speakers into their respective categories. Support vector machine classification method has been successfully used to perform the assessment of the dysarthria severity level. Keywords: Keywordsdysarthria, rhythm, pairwise variability index, acoustical analysis, Nemours database, support vector machine, SVM, classification, discriminant analysis

Résumé: An OFDM system is combined with multiple-input multiple-output (MIMO) in order to increase the diversity gain and system capacity over the time variant frequency-selective channels. However, a major drawback of MIMO-OFDM system is that the transmitted signals on different antennas might exhibit high peak-to-average power ratio (PAPR).In this paper, we present a PAPR analysis reduction of space-timeblock-coded (STBC) MIMO-OFDM system for 4G wirelessnetworks. Several techniques have been used to reduce the PAPR of the (STBC) MIMOOFDM system: clipping and filtering, partial transmit sequence (PTS) and selected mapping (SLM). Simulation results show that clipping and filtering provides a better PAPR reduction than the others methods and only SLM technique conserve the PAPR reduction in reception part of signal

Résumé: An energy-efficient discrete cosine transform (DCT) is proposed. It can be used in image compression in wireless visual sensor networks. It is a combination of the recently proposed block discrete cosine transform and a pruned approach. Thus, the computational complexity is reduced significantly. The experimental results show that the proposed transform leads to a significant reduction in computation time at the target sensor node. This will consequently result in a saving in energy consumption. The proposed DCT is very suitable for implementation in wireless sensor networks powered by batteries.

Résumé: As any existing video codec, multiview video coding requires a highly efficient compression as well as a faster random access to any given view. Such efficiency for the coding approach can be achieved by exploiting both the temporal and the inter-view correlation. The temporal and view random access are improved by the increase of intracoded pictures (I pictures), which requires more bit rate. In this research study, we propose and evaluate an improved inter-view prediction structure mainly based on reducing the complexity of view random access at any instance T n and improving the inter-view correlation for P-view. These improvements can be guaranteed by the appropriate choice of the position of I-view, the type of the different views, and the number of each frame type to use. Compared to the prediction scheme, which is based on the hierarchical B pictures structure and used for joint multiview video model, experimental results have shown that the proposed prediction structure provides better bit-rate performance of up to 6.81% with a similar compression quality measured in peak signal-to-noise ratio and significantly improves view random access by up to 25%.

Résumé: This paper presents a methodology which can be used to implement any decimator symmetric/antisymmetric (S/A) finite impulse response (FIR) filter. Two varieties are developed: a classic distributed arithmetic (CDA) based and a modified distributed arithmetic (MDA) based one. Both exploit the polyphase structure and the symmetry/antisymmetry of the filter and are evaluated in terms of area efficiency, speed and power consumption. The choice of the algorithm depends on the performance metrics targeted. The methodology has been applied to implement the filter bank CDF9/7 which constitutes a one dimensional (1D) and one level discrete wavelet transform (DWT). The filter bank also known as the bior4.4 biorthogonal wavelets is recommended by the JPEG2000 standard for lossy compression of images and video. The architecture has been implemented on an Altera field programmable gate array (FPGA) and the simulations run in Matlab, Modelsim and Altera Quartus II. The results prove the efficiency of the algorithms and show the tradeoff between the area occupied, the throughput and the power consumption.

Résumé: This paper reports the results of acoustic investigation based on rhythmic classifications of speech from duration measurements carried out to distinguish dysarthric speech from healthy speech. The Nemours database of American dysarthric speakers is used throughout experiments conducted for this study. The speakers are eleven young adult males with dysarthria caused by cerebral palsy (CP) or head trauma (HT) and one non-dysarthric adult male. Eight different sentences for each speaker were segmented manually to vocalic and intervocalic segmentation (176 sentences). Seventy-four different sentences for each speaker were automatically segmented to voiced and non-voiced intervals (1628 sentences). A two-parameters classification related to rhythm metrics was used to determine the most relevant measures investigated through bi-dimensional representations. Results show the relevance of rhythm metrics to distinguish healthy speech from dysarthrias and to discriminate the levels of dysarthria severity. The majority of parameters was more than 54% successful in classifying speech into its appropriate group (90% for the dysarthric patient classification in the feature space (%V, ΔV)). The results were not significant for voiced and unvoiced intervals relatively to the vocalic and intervocalic intervals (the highest recognition rates were: 62.98 and 90.30% for dysarthric patient and healthy control classification respectively in the feature space (ΔDNV, %DV)). Keywords: Dysarthria; Rhythm; Pairwise variability index; Acoustical analysis; Timing; Nemours database; Dysarthric severity

Résumé: In this paper, we introduce an algorithm for motion estimation. It combines complex wavelet decomposition and a fast motion estimation method based on affine model. The principle of wavelet transform is to decompose hierarchically the input image into a series of successively lower resolution reference images and detail images which contain the information needed to be reconstructed back to the next higher resolution level. The motion estimation determines the velocity field between two successive images. This phase can be extracted from this measure descriptive information of the sequence. Motion Estimation (ME) is an important part of any video compression system, since it can achieve significant compression by exploiting the temporal redundancy existing in a video sequence. This paper described a method from calculating the optical flow of an image sequence based on complex wavelet transform. It consists to project the optical flow vectors on a basis of complex-valued wavelets. Thus, we add an additional assumption on the shape of the velocity field that we want to find, which is the affinity of the optical flow. The two-dimensional affine motion model is used to formulate the optical flow problem by coarse resolution simultaneously coarse-and-fine, beside the traditional approach by coarse-to-fine, to avoid the error propagation during the decomposition of coarse level to fine level. This method opens the way for a quick and low-cost computing optical flow. Keywords: Motion estimationComplex waveletFast two frame algorithmCoarse and fine model

Résumé: In this paper, we propose an image compression algorithm that uses a hybrid transform and an improved modified set partitioning in hierarchical trees (SPIHT) coding algorithm. The proposed transform uses the subband discrete cosine transform to decompose the image into multiresolution subbands where the discrete wavelet transform is then used to code the low frequencies. Then, we use the SPIHT coding method to code the transformed coefficients. For the SPIHT algorithm, we have proposed a method to reduce the distortion introduced by the SPIHT technique between the original and reconstructed images. The obtained results show the efficiency of the proposed hybrid method in terms of peak signal-to-noise ratio and visual quality.

Résumé: For the purpose of biometric applications, we explore in this paper a new robust approach to characterizing palmprint features. Instead of processing the acquired image in the spatial domain, the proposed technique extracts palmprint features using Radon transform and a geometric Delaunay triangulation jointly. In such a process, Radon transform enables the extraction of directional characteristics from the palm of the hand. Afterwards, the most significant information is structured using Delaunay triangulation, thus providing a specific palmprint signature. In order to compare the uniqueness as well as the stability of the palmprint signature, Hausdorff distance has been used as a criterion of similarity. As will be shown in this paper, the palmprint signature is very robust even when considering a low Signal-to-Noise Ratio (SNR). Promising results are obtained from a local database containing 200 palmprint images. This technique is mainly appropriate for authentication applications

Résumé: In this paper, the use of finite Gaussian mixture m odal (GMM) based Expectation Maximization (EM) estimated algorithm for score lev el data fusion is proposed. Automated biometric systems for human identification measure a “signature” of the human body, compare the resulting characteristic to a database, and render an application dependent decision. These biometric systems for personal auth entication and identification are based upon physiological or behavioral features which are typically distinctive, Multi-biometric systems, which consolidate information from multipl e biometric sources, are gaining popularity because they are able to overcome limita tions such as non-universality, noisy sensor data, large intra-user variations and suscep tibility to spoof attacks that are commonly encountered in mono modal biometric systems. Simula tion show that finite mixture modal (GMM) is quite effective in modelling the genuine a nd impostor score densities, fusion based the resulting density estimates achieves a signific ant performance on eNTERFACE 2005 multi-biometric database based on face and speech modalities. Keywords: Biometry, Multi-Modal, Authentication, Face Recogni tion, Speaker Verification, data Fusion, Adaptive Bayesian decisi on, GMM & EM..

Résumé: This paper presents a comparative analysis of the performance of three estimation algorithms: Expectation Maximization (EM), Greedy EM Algorithm (GEM) and Figueiredo-Jain Algorithm (FJ) - based on the Gaussian mixture models (GMMs) for signature biometrics verification. The simulation results have shown significant performance achievements. The test performance of EER=5.49 % for "EM", EER=5.04 % for "GEM" and EER=5.00 % for "FJ", shows that the behavioral information scheme of signature biometrics is robust and has a discriminating power, which can be explored for identity authentication.

Résumé: In this work, we first give a general method which can get a large class of balanced Boolean functions with reasonably high nonlinearity, larger than that obtained by Lobanov. Then, we study the secondary construction of Boolean functions without extending their number of variables, introduced recently by Carlet. This gives interesting cryptographic properties in terms of balancedness, nonlinearity and algebraic immunity. We conclude the paper by proving that the algebraic immunity of the constructed functions is better than among of the starting functions.

Résumé: ABSTRACT In this paper, we present the application of adaptive wavelet in lossy image compression. The construction of this adaptive wavelet is realised by the use of lifting scheme. In our application the lifting scheme is composed by an adaptive update lifting step and a fixed prediction lifting step. Finally, experiments with The EBCOT (Embedded Block Coding with Optimized Truncations) algorithm applied on synthetic and real images are reported.

Résumé: tate reconstruction approach is very useful for sensor fault isolation, reconstruction of faulty measurement and the determination of the number of components retained in the principal components analysis (PCA) model. An extension of this approach based on a Nonlinear PCA (NLPCA) model is described in this paper. The NLPCA model is obtained using five layer neural network. A simulation example is given to show the performances of the proposed approach. Keywords: Fault detection and isolationreconstructionnonlinear PCA (NLPCA)neural networks

Résumé: We propose a new approach to image compression based on the principle of Shapiro's embedded zero-tree wavelet (EZW) algorithm. Our approach, the efficient EZW (E-EZW), uses six symbols instead of the four used in the original Shapiro's algorithm to minimize the redundant symbols, and optimizes the coding by binary regrouping of the information. This approach can produce results that have a significant improvement over the peak signal-to-noise ratio and compression ratio obtained by Shapiro without affecting the computing time. These results are also comparable to those obtained using the set partitioning in hierarchical trees (SPIHT), set partitioning embedded block (SPECK), and JPEG2000 algorithms.

Résumé: This paper reports the effect of neutron irradiation defects on electrical properties of n-type FZ-silicon via the measurements of electrical resistivity. For this purpose, FZ-silicon single crystal was irradiated with neutron fluences ranging between 1.54 × 1016 and 2.5 × 1018 cm−2. The samples irradiated at F1 = 1.54 × 1016 cm−2, F2 = 7.43 × 1017 cm−2 and F3 = 2.5 × 1018 cm−2 were isochronally annealed from room temperature up to 750 °C. It is found that, for fluences ranging respectively from 1.54 × 1016 cm−2 to 1.23 × 1017 cm−2 (stage I) and from 3.09 × 1017 cm−2 to 2.5 × 1018 cm−2 (stage II) the resistivity is linearly related to the neutron fluence with two different slopes. The annealing temperature dependence on the electrical resistivity fits well the relationρ0 exp(−CT), where C is a constant depending on the neutron fluence and ρ0 is approximately equal to the resistivity after irradiation. For annealing temperatures higher than 550 °C, we have found that the resistivity is a decreasing function with respect to the neutron fluence and the transmutation-doped phosphorus atoms become electrically active.

Résumé: The work suggested in this research consists of the detection of a three-phase short circuit in a synchronous permanent magnets machine supplied with a three levels inverter of voltage PWM, by the means of the spectral analysis of the stator currents. The practice showed that at the time of a short circuit it is primarily the stator resistance which considerably will increase. For that we establish the model of the operational engine first of all. Then we will give the model of the engine weakening and will treat the results of simulation for finally comparing the spectra of harmonics of the currents of the two models. The early knowledge of such a defect starting from a simple spectrogram makes it possible to avoid the total dysfunction of the machine.

Résumé: In this paper, a new approach on nondestructive evaluation by ultrasonic signal is proposed. This work is based on a qualitative analysis (parametric method) of rock acoustic characteristics by autoregressive (AR) spectral estimation of ultrasonic signals transmitted through materials (rock samples). The longitudinal waves (P) are chosen in the experiment analysis (the case of transversal waves (5) will be tested in other studies). The (AR) model parameters determined by predictor algorithms give a very helpful information which can be used to realise good and accurate rock diagnosis.

Résumé: In this paper, a new approach on nondestructive evaluation by ultrasonic signal is proposed. This work is based on a qualitative analysis (parametric method) of rock acoustic characteristics by autoregressive (AR) spectral estimation of ultrasonic signals transmitted through materials (rock samples). The longitudinal waves (P) are chosen in the experiment analysis (the case of transversal waves (5) will be tested in other studies). The (AR) model parameters determined by predictor algorithms give a very helpful information which can be used to realise good and accurate rock diagnosis.

Résumé: In order to determine the intrinsic (or elastic) energetic shape of Auger peaks two contributions to the experimental spectrum are defined: one elastic and the other inelastic. These contributions are analytically expressed in terms of the elastic and inelastic mean free paths and of the differential inelastic scattering cross-section. Next we show how to directly subtract the inelastic contribution. It is also shown that, under specific and often verified conditions, the deconvolution of the experimental spectrum (by the backscattering spectrum corresponding to an energy of incident electrons which matches Auger energy) makes it possible to determine the desired intrinsic energetic distribution. Both methods were applied to the same experimental spectra and the obtained results were compared. Compared to the “Sickafus law”, the method used allows a better account of the secondary electron contribution to the background. To perform the deconvolution, an improved Van-Cittert iterative method has been used.

Résumé: Méthode utilisée ensuite sur un alliage cuivre-aluminium contenant seulement 1% de cuivre

Résumé: Etude de la résolution spatiale en analyse Auger en tenant compte de la contribution du faisceau direct et des électrons rétrodiffusés. L'influence de la taille de sonde sur la distribution spatiale des électrons Auger créés par les électrons rétrodiffusés a été simulée

Résumé: Etude de la résolution spatiale en analyse Auger en tenant compte de la contribution du faisceau direct et des électrons rétrodiffusés. L'influence de la taille de sonde sur la distribution spatiale des électrons Auger créés par les électrons rétrodiffusés a été simulée

Résumé: An effective scheme for coding of arbitrarily shaped color visual objects is presented. Based on Set Partitioning In Hierarchical Trees (SPIHT) algorithm, the proposed algorithm employs the Shape-Adaptive Discrete Cosine Transform (SA-DCT) in the place of the Shape-Adaptive Discrete Wavelet Transform (SA-DWT) to code texture coefficients by using mask pixels in order to create an embedded code that allows for fine-grained rate-distortion scalability. The result of the proposed procedure is a multi-resolution, progressive reconstruction of the binary shape mask as well as the texture. This algorithm is compared with wavelet-based embedded coding of the object texture. Objective and subjective simulation results show that the proposed scheme has excellent rate-distortion performance.

Résumé: In this paper, we introduce a compression algorithm using wavelet transform. The principle of wavelet transform is to decompose hierarchically the input image into a series of successively lower resolution reference images and detail images which contain the information needed to be reconstructed back to the next higher resolution level . The histogram of image sub-bands provides us with information on the distribution of the coefficient values in this subimage. The sub-band images resulting from wavelet transform are not of equal significance. Some sub-bands contain more information than others. The total number of available bits describing an image is however inevitably limited. Therefore, it is desirable to allocate more bits to those sub-bands images which can be coded more accurately than others. The objective of a such bit allocation method is to optimize the overall coder performance and minimize the quantization error. In determining which wavelet filter is to be used for image compression, some of the properties considered are vanishing moments. The phase non-linearity of the filter can cause severe degradation in the subjective quality of an image. It is related to the symmetry of the filter coefficients. The wavelet transform is implemented using a linear-phase Biorthogonal filter with four levels of decomposition. For this study, we use a scalar quantization with uniform threshold quantizers. The quantization method is PCM (pulse coded modulation) for the coefficients in all high-pass sub-bands. The coefficients of low-pass sub-bands are DPCM (Differential PCM) quantized per region.

Résumé: Dans ce travail, nous nous intéressons à la compression d’ images biomédicales fixes par différents types de la transformée en ondelettes discrètes, associés à différents algorithmes de quantification vectorielle et de codage entropique. Ce type de compression nous a permis de déterminer la qualité des images reconstruites (PSNR) et le taux de compression (TC) correspondants selon le type de l’ondele tte et les algorithmes de QV et de codage entropique utilisés. Une étude comparative a été menée dans le but de dé terminer les méthodes conduisant aux meilleurs résultats possibles.

Résumé: In this paper, a new real-time approach for audio recognition of waterbird species in noisy environments, based on a Texas Instruments DSP, i.e. TMS320C6713 is proposed. For noise estimation in noisy water bird's sound, a tonal region detector (TRD) using a sigmoid function is introduced. This method offers flexibility since the slope and the mean of the sigmoid function can be adapted autonomously for a better trade-off between noise overvaluation and undervaluation. Then, the features Mel Frequency Cepstral Coefficients post processed by Spectral Subtraction (MFCC-SS) were extracted for classification using Support Vector Machine classifier. A development of the Simulink analysis models of classic MFCC and MFCC-SS is described. The audio recognition system is implemented in real time by loading the created models in DSP board, after being converted to target C code using Code Composer Studio. Experimental results demonstrate that the proposed TRD-MFCC-SS feature is highly effective and performs satisfactorily compared to conventional MFCC feature, especially in complex environment.

Résumé: Wetlands are home to an impressive number of fauna and flora, some of which are threatened with extinction. There is a large number of wetlands around the world, among which the Pantanal is surely the major one. It covers a total area of 150,000 km2 across three countries of South America (Brazil, Bolivia, and Paraguay). It is home to very rich and numerous fauna and flora, including more than 650 bird species. There are other important wetlands in the world, such as Camargue (France), the Everglades (United States), or Okavango (Botswana). Algeria has more than 250 wetlands, of which 50 are classified internationally for their importance and their ecological role. They are also privileged places for tens of thousands of waterbirds of different species to overwinter or make a temporary halt. Some of these species are threatened with extinction according to the latest classification of the International Union for Conservation of Nature (IUCN) such as the White-headed Duck (Oxyura leucocephala) and the Ferruginous Duck (Aythya nyroca). This gives Algeria's wetlands a great importance and requires special attention. In this chapter, we propose a WSN-based monitoring of these birds in their natural habitat. The idea is to identify and recognize different bird species instantly from the detection and processing of their vocalizations (call and/or song) by wireless sensor nodes. Sensor networks are proving to be ideal platforms for recording and processing such data because of their characteristics compliance with the requirements of the project such as energy independence, the average financial cost, wide geographic coverage, and the preservation of the environment. However, the proposed monitoring method, based on the WSN, must meet two key challenges. First, immunity to environmental noises that are inevitably present in all recorded bird sounds, and second, a reduced computational complexity leading to an energy consumption saving of the wireless sensor nodes and thus increasing their lifetime.

Résumé: Recent video coding standards such as Advanced Video Coding (AVC/H.264) and High Efficiency Video Coding (HEVC/H.265) provide extension profiles allowing for the exploitation of the inter-view resemblances that exist in multiview video for better compression efficiency. This chapter presents an overview of the multiview coding concepts while focusing on the extended profiles of MVC and MV-HEVC, respectively. Also, experiments have been conducted comparing MVC and MV-HEVC with regard to their compression efficiency. Multiview video sequences with different resolutions and textures were subject to the tests. The reported results show that MV-HEVC outperforms MVC in terms of bit rate saving, yielding a gain of over 70% for a selected video sequence.

Résumé: The latest multi-view video coding generation MV-HEVC exploited effectively the interview redundant information in multi-view videos and provided high compression efficiency. However, increasing the distance between cameras during multi-view videos acquisition creates an inter-view disparity. This latter, influences the inter-view similarities and affects the disparity estimation which is requiring an individual estimation for each coding unit. This process increases the computational complexity of the MV-HEVC encoder. In this paper, we propose an improved disparity estimation by eliminating the offset between views during inter-view encoding. This elimination of the offset leads to the decrease of the search window in disparity estimation process according to the coding units size. For reliable motion estimation, an earlier decision of coding units splitting (EDCS) is also proposed. Experimental results show that the proposed approach roughly reduces the encoding time from 23.31 % to 44.35 % on average with negligible performance degradation.

Résumé: The evolution of the Internet and wireless communications led real-time applications with high and ultra-high-definition video to flourish. However, due to the fact that video is a bandwidth-greedy media stream, numerous codecs such as AV1 and Versatile Video Coding (VVC) that implement different techniques, have emerged. Indeed, investigating the performance of different compression standards is crucial to select the appropriate codec for a specific application. This paper provides a comparative assessment of the compression efficiency for the above-mentioned standards according to their prominent predecessors VP9, HEVC and H.264. Seven encoder implementations were used for benchmarking AV1, VP9, VVC, HEVC and H.264. i.e. AV1, VP9, VTM, (HM and x265), as well as (JM and x264) respectively. The low-dealy configuration pattern was utilized as it provides the lowest delay in contrast to random-access and all-Intra configurations. Indeed, to address real-time mobile applications, class C video sequences that are very sought after in many applications such as video surveillance or road safety were used for the conducted experiments. The obtained results show that each newly born codec outperforms its predecessor in terms of coding efficiency with an increase in the computational complexity. VVC achieves the highest compression performance in comparison to its predecessors even with the optimized version (x265, x264) with a noticeable increase in encoding run time, while VP9 provides average bit-rate overhead according to its successor AV1 for the industrial standards.

Résumé: In this paper we propose a novel efficient low-complexity Discrete Cosine Transform (DCT) approximation using only 14 additions. The proposed transformation is based upon a pruned version of the Signed DCT (SDCT). Associated with a JPEG compression chain, this DCT approximation ensures a very good rate-distortion compromise, a very low computational complexity and a significant compatibility with the exact DCT. The proposed transformation is very suitable for the limited energy resource wireless multimedia sensor networks (WMSNs) requiring very low bitrates and other embedded systems. Simulation results show the superiority of the proposed fast algorithm compared to any existing pruned DCT approximations.

Résumé: The Discrete cosine transform (DCT) plays an important role in image and video coding. It is used as the first step in lossy/lossless image compression. Consequently, many algorithms have been developed to efficiently compute it. In the image compression community, an alternative to the DCT, known as the discrete Tchebichef transform (DTT), is increasingly used. It possesses an excellent energy compaction property and a low implementation complexity. However, the DTT has not been investigated in the compression of color images. In this paper, the theory behind the DTT and its pruned version (PDTT) are investigated and a lossy compression method of color images based on the DTT is proposed. Experimental results demonstrate that the proposed method exhibits almost the same performance as the DCT-based technique, in terms of objective measures such as the PSNR and SSIM while reducing the arithmetic complexity by more than 37%.

Résumé: High-quality delivery of compressed video over vehicular networks is a very challenging task. In this paper, we propose a cross-layer mechanism dedicated to high-efficiency video coding (HEVC) low delay temporal prediction structure in order to improve the video transmission through vehicular ad-hoc networks (VANETs). Compressed video is obtained thanks to the state-of-the-art HEVC video coder which exploits new efficient coding tools and structures. Then, a mechanism is proposed which accounts for the type of each frame as well as the network traffic load state. Furthermore, the proposed mechanism guarantees better resources exploitation of the VANET oriented IEEE 802.11p standard. Indeed, the proposed system exploits the medium access control (MAC) layer access categories (ACs) that are not dedicated to video transmission. The results obtained from realistic VANET simulations demonstrate that the proposed cross-layer mechanism achieves significant improvements compared to the actual standard in terms of packet delivery rate as well as PSNR gains at the reception.

Résumé: Recently, video streaming has attracted much attention and interest due to its capability to process and transmit large data. We propose a quality of experience (QoE) model relying on high efficiency video coding (HEVC) encoder adaptation scheme, in turn based on the multiple description coding (MDC) for video streaming. The main contributions of the paper are (1) a performance evaluation of the new and emerging video coding standard HEVC/H.265, which is based on the variation of quantization parameter (QP) values depending on different video contents to deduce their influence on the sequence to be transmitted, (2) QoE support multimedia applications in wireless networks are investigated, so we inspect the packet loss impact on the QoE of transmitted video sequences, (3) HEVC encoder parameter adaptation scheme based on MDC is modeled with the encoder parameter and objective QoE model. A comparative study revealed that the proposed MDC approach is effective for improving the transmission with a peak signal-to-noise ratio (PSNR) gain of about 2 to 3 dB. Results show that a good choice of QP value can compensate for transmission channel effects and improve received video quality, although HEVC/H.265 is also sensitive to packet loss. The obtained results show the efficiency of our proposed method in terms of PSNR and mean-opinion-score.

Résumé: In this paper, an efficient video transmission technique is proposed for vehicular communications using the TCP Friendly Rate Control (TFRC) transport protocol. It allows video adaptation at the transmitter side through a channel state estimation feedback. Simulation results demonstrate the effectiveness of the method in terms of received video quality.

Résumé: The integration of biometric technologies into surveillance systems is a major step milestone to improve the automation process in order to recognize criminal offenders and track them across different places. The suitability of gait biometrics for surveillance applications emerges from the fact that the walking pattern can be captured and perceived from a distance even with poor resolution video as opposed to other biometric modalities which their performance deteriorates in surveillance scenarios. In this research article, we explore the use of gait biometrics using features derived from the lower part of legs for people identification. The gait features are extracted from both spatial and temporal domains. These features are mainly estimated from the Shapes of the Gaps between the Lower Limbs (SGLLs) when walking using the Hu's Invariant Moments. The spatial features are those of a single SGLL extracted from one image and considered as local characteristics. Comparative analysis is conducted against well-established methods. The attained results confirm that people identification using gait features extracted from the lower limbs is still perceivable with better recognition rates even under the influence of different covariate factors.

Résumé: In this paper, we propose a multimodal palmprint verification system based on two bands (gray scale and near infrared) using a progressive image compression through the famous Set Partitioning In Hierarchical Trees (SPIHT) coder. Each palmprint image is compressed an decompressed at 0.5 bit per pixel (bpp). By exploiting the progressiveness of the SPIHT algorithm, we decompressed the palmprint images in two bitrates. We obtained three images, an image at 0.25 bpp which represents the image approximation, the other at 0.25 bpp which represents the edges or details, and the third is the global of the both preceding. After analyzing the extracted features from different bands, we propose a score-level fusion scheme to integrate the multimodal information. The palmprint verification experiments demonstrated the superiority of multibands fusion to each single band, which results in higher verification accuracy.

Résumé: In this paper, we propose an efficient multi-spectral palmprint identification system. For that, we compressed each palmprint band at 0.5 bit per pixel (bpp) by a progressive image compression algorithm where we used the famous Set Partitioning In Hierarchical Trees (SPIHT ) coder. By exploiting the progressiveness of this algorithm, we obtained three images decompressed in several times with different resolutions. One image decompressed at 0.25 bpp which represents the image approximation, the other at 0.25 bpp which represents the edges or details and the third is the global of them. So, we applied a binarization scheme and the Gabor algorithm for the last two images respectively for modeling. Subsequently, all bands are integrated in order to construct an efficient multimodal identification system based on matching score level fusion. Finally, Experimental results show that our proposed scheme yields excellent performance for identifying palmprints.

Résumé: The view random access is one of most important requirements for video compression algorithms. The acceleration of view random access can be obtained by the appropriate choice of the used prediction structure. This acceleration is achieved by several constraints. The choice of the prediction structure and the evaluation method can be among the most important constraints. In this research study, we propose an improved inter-view prediction structure that is mainly based on the use of successive B-views. Further, we describe a novel evaluation method in order to assess the efficiency of the proposed approach. This method is based on two metrics which are; the number of reference images needed to decode a given image, in addition to the maximum number of reference frames to consult a given image. Experimental results for the view random access have shown significant improvement that exceeds 37.5% with respect to the IBP prediction structure of the reference model JMVM. In the same way, the improved bit-rate of the proposed approach can reach 7.11 % with respect to the IBP structure for a similar video quality.

Résumé: In this paper, we propose to detect and localize the broken bar faults in multi-winding induction motor using Motor current signature (MCSA) combined to Support Vector Machine (SVM). The analysis of stator currents in the frequency domain is the most commonly used method, because induction machine faults often generates particular frequency components in the stator current spectrum. In order to obtain a more robust diagnosis, we propose to classify the feature vectors extracted from the magnitude of spectral analysis using multi-class SVM to discriminate the state of the motor. Finally, in order to validate our proposed approach, we simulated the multi-winding induction motor under Matlab software. Promising results were obtained, which confirms the validity of the proposed approach.

Résumé: The performance of a biometric system is based primarily on the quality of physical or behavioral biometric used for a robust and an accurate authentication/identification of an individual. To improve the performance and the robustness of the system, multispectral palmprint images were employed to acquire more discriminative information. In this paper, we introduce a novel multispectral recognition method. In this context, we propose the fusion of palmprint and palm vein features to increase the accuracy of the biometric person recognition. The proposed approach is based on statistical study and energy distribution analysis of Finite Ridgelet transform coefficients, involving so low computation complexity. For multispectral palmprint images recognition, we tested the performance of three classifiers: K nearest neighbor (KNN), Support Vector Machine (SVM) and `One-Against-One' multi-class SVM (OAO-SVM) with RBF kernel using 6-folders cross-validation to assess the generalization capability of the proposed biometric system. The validation of our results is performed on multispectral palmprint images of CASIA database.

Résumé: Images segmentation constitutes a crucial task in magnetic resonance images analysis by automating and facilitating isolation of anatomical structures and other regions of interest. The purpose of this paper is to present a state of the art of current approaches dealing with anatomical magnetic resonance images segmentation. Emphasis will be placed on revealing the advantages and disadvantages of these methods for medical imaging applications.

Résumé: Optical flow computation is an important and challenging problem in the motion analysis of images sequence. It is a difficult and computationally expensive task and is an ill-posed problem, which expresses itself as the aperture problem. However, optical flow vectors or motion can be estimated by differential techniques using regularization methods; in which additional constraints functions are introduced. In this work we propose to improve differential methods for optical flow estimation by including colour information as constraints functions in the optimization process using a simple matrix inversion. The proposed technique has shown encouraging results.

Résumé: A new method for the subspace-based direction of arrival (DOA) estimation procedure without eigenvector computation is proposed. From the residual vectors of the conjugate gradient (CG) method which form a Krylov subspace basis, we build a test spectrum for DOA estimation. This approach is based on the same recently developed procedure which uses a non-eigenvector basis derived from the auxiliary vectors (AV). The AV basis calculation algorithm is replaced by the residual vectors of the CG algorithm. The initial conditions of the CG algorithm start with the linear transformation of the array response search vector by the input covariance matrix. Then, successive orthogonal gradient vectors are derived to form a basis of the signal subspace. The proposed CG-based method outperforms its counterparts in term of resolution of closely spaced-sources with a small number of snapshots and a low signal-to-noise ratio (SNR).

Résumé: A new method for the subspace-based direction of arrival (DOA) estimation procedure without eigenvector computation is proposed. From the residual vectors of the conjugate gradient (CG) method which form a Krylov subspace basis, we build a test spectrum for DOA estimation. This approach is based on the same recently developed procedure which uses a non-eigenvector basis derived from the auxiliary vectors (AV). The AV basis calculation algorithm is replaced by the residual vectors of the CG algorithm. The initial conditions of the CG algorithm start with the linear transformation of the array response search vector by the input covariance matrix. Then, successive orthogonal gradient vectors are derived to form a basis of the signal subspace. The proposed CG-based method outperforms its counterparts in term of resolution of closely spaced-sources with a small number of snapshots and a low signal-to-noise ratio (SNR)

Résumé: The aim of our work is to perform an automated tool for brain MRI tissues quantification. The method we develop is based on MRI intensity stochastic analysis. By the use of the Gaussian mixture model for these intensities, we estimate MRI tissues parameters with a combination of the expectation-maximization algorithm and the Markov random field model witch provide contextual constraints that improve the classification of image pixels into three classes of tissue: white matter, grey matter and cerebro-spinal fluid. The automated model based algorithm is also extended to take in account an important MRI artefact: the bias field caused by electromagnetic field inhomogeneities. The resulting automated MRI analysis method simultaneously corrects from MR field inhomogeneities, estimates tissue classes distribution parameters, classifies the image and detects multiple sclerosis lesions when treated images present this pathology. We validate our method on simulated data then on real MRI scans.

Résumé: In this paper, a new approach of images coding by Shapiro algorithm (embedded zerotree wavelet algorithm or EZW) is proposed. This approach, the modified EZW (MEZW), distributes entropy differently than Shapiro's and also optimizes the coding. This can produce results that are a significant improvement on the PSNR and compression ratio obtained by Shapiro, without affecting the computing time. These results are also comparable with those obtained using the SPIHT and SPECK algorithms. The EZW, Spiht or Speck algorithms are based on the wavelet transform. The principle of wavelet transform is to decompose hierarchically the input image into a series of successively lower resolution reference images and detail images which contain the information needed to be reconstructed back to the next higher resolution level. The subband images resulting from wavelet transform are not of equal significance. Some subbands contain more information than others (example the baseband subband)

Résumé: The aim of our work is to perform an automated tool for brain MRI tissues quantification. The method we develop is based on MRI intensity stochastic analysis. By the use of the Gaussian mixture model for these intensities, we estimate MRI tissues parameters with a combination of the expectation-maximization algorithm and the Markov random field model witch provide contextual constraints that improve the classification of image pixels into three classes of tissue: white matter, grey matter and cerebro-spinal fluid. The automated model based algorithm is also extended to take in account an important MRI artefact: the bias field caused by electromagnetic field inhomogeneities. The resulting automated MRI analysis method simultaneously corrects from MR field inhomogeneities, estimates tissue classes distribution parameters, classifies the image and detects multiple sclerosis lesions when treated images present this pathology. We validate our method on simulated data then on real MRI scans

Résumé: In this paper, the filter's effect in image compression by wavelet transform is studied. The principle of wavelet transform is to decompose hierarchically the input image into a series of successively lower resolution reference images. A scalar quantization with uniform threshold quantizers is used. The biorthogonal filters provide the good result in image compression. The reconstructed image quality is evaluated by mean-squared error (MSE) and peak signal to noise ratio (PSNR).

Résumé: In this paper, the estimation of the disturbing frequencies injected by nonlinear load and the elimination of the parasitic harmonics in the electrical supply network is analyzed. The measurement in this paper is accompanied by colored additive noise. The noise is modeled by a filter AR of a finite order. The results of estimation and identification obtained are compared with Pizarenko, Prony, Music and SVD (singular value decomposition) techniques.

Résumé: In this paper, we propose a new method for the adaptation of SPIHT algorithm (set partitioning in hierarchical trees) to the color images. The goal is to effectively exploit the space and interspectral redundancies of the three components of the color. A slight modification of the "parent-children" relation, defined in SPIHT, was introduced, taking account of the interspectral correlation and the energy distribution of the image through the three components.

Résumé: In this paper, a new approach on nondestructive evaluation by ultrasonic signal is proposed. The work is based on a parametric spectral analysis with linear prediction for a sharp and precise analysis. The paper contains a description of quantitative analysis method of rocks acoustic properties and a description of a qualitative analysis method based on autoregressive (AR) spectral estimation. The diagnosis done by the prediction error when the spectra (or power spectral density) present quick changes, which indicate the existence of microcracks and pores inside the specimen.

Résumé: We address the problem of the estimation of real sinusoids in colored noise using methods based on the forth-order cumulants. In this paper we focus our attention on the resolution problem. We propose to compute the forth order cumulant matrix in order to solve the estimation problem of amplitudes and frequencies from data blurred with colored noise of unknown power spectral density among the many that allow the reconstruction. For such cases, the methods based on the forth order cumulants are asymptotically efficient and have excellent performances. A simulated example is also presented to show the excellent performance in resolution of approach.

Résumé: We introduce a compression algorithm using the wavelet transform. The principle of wavelet transform is to decompose hierarchically the input image into a series of successively lower resolution reference images and detail images that contain the information needed to be reconstructed back to the next higher resolution level. The histogram of image sub-bands provides information on the distribution of the coefficient values in this sub image. The sub band images resulting from wavelet transform are not equal significance. Some sub-bands contain more information than others. The total number of available bits describe an image is however, inevitably limited. Therefore, it is desirable to allocate more bits to those sub-band images and can be coded more accurately than others. The objective of such bit allocation method is to optimize the overall coder performance and minimize the quantization error. In determining which wavelet filter is to be used for image compression, some of the properties including vanishing moments are considered. The wavelet transform is implemented using a linear-phase biorthogonal filter (16,4) with 4 level of decomposition. For this study, we used a scalar quantization with uniform threshold quantizers. The quantization method is pulse code modulation (PCM) for the coefficients in all high-pass sub-bands. The coefficients of low-pass sub-band are differential PCM quantized per region.