La radio astronomie étudie les sources cosmiques au travers de leur rayonnement dans le domaine radio. Les astronomes, utilisateurs passifs du spectre électromagnétique, ont à faire face à une pollution radio de plus en plus importante. Cette thèse s’intéresse particulièrement aux interférences radio d’origine humaine (RFI), et comment les observations radio astronomiques peuvent être réalisées en bandes de fréquences non-protégées. Les approches classiques consistent à contrôler les paramètres statistiques d’une observation. Une fois détectées, les données polluées sont retirées avant post-traitement. En plus d’autres avantages techniques par rapport aux radiotélescopes paraboliques classiques, les réseaux d’antennes offrent une information spatiale lors d’une observation astronomique. La diversité spatiale entre source cosmique d’intérêt (SCOI) et RFI peut être exploitée pour développer des traitements spatiaux d’interférences. Après la formulation d’un module de données multidimensionnel, une technique de soustraction de sous espace RFI est introduite. Cette technique consiste à soustraire la contribution des RFI aux données d’une observation. La projection orthogonale a déjà été considérée auparavant. Cependant, l’orthogonalité requise entre CSOI et RFI pour retrouver une source d’intérêt non biaisée ne peut vraisemblablement pas être satisfaite. Une approche basée sur une projection oblique est introduite afin de pallier à cette condition. Les techniques de projections sont comparées aux techniques classiques de beamforming en termes de réjection de l’interférence et de récupération de la source d’intérêt. Le sous-espace RFI est inconnu de manière générale et se doit d’être estimé. Plusieurs techniques permettant cette estimation, basées sur des propriétés statistiques des RFI et sources cosmiques, sont également présentées et comparées. Les différentes techniques ont été appliquées à des données astronomiques délivrées par le radio télescope Européen LOFAR. Enfin, une implémentation d’un algorithme de traitement spatial d’interférences sur le démonstrateur EMBRACE est présenté. / Radio astronomy studies cosmic sources through their radio emissions. As passive users, astronomers have to deal with an increasingly corrupted radio spectrum. The research presented here focuses on man-made Radio Frequency Interference (RFI), and how astronomical observations can be performed in non-protected frequency bands. Traditional approaches consist in monitoring radio telescopes output data through statistical parameters. Once detected, the corrupted data is removed before further processing. Besides other technical advantages compared to single dish radio telescopes, antenna arrays provide spatial information about astronomical observations. The spatial diversity between cosmic sources-of-interest (CSOI) and RFI can be exploited to develop spatial RFI processing. After formulating a multidimensional radio astronomical data model, an interference subspace subtraction technique is introduced. This approach consists in subtracting RFI contributions from antenna array radio telescopes data. Orthogonal projection applied to astronomical observation vector spaces has already been considered by the past. The orthogonality between RFI and CSOI subspaces is required to recover the CSOI without bias. In order to avoid this latter requirement, an oblique projection approach is here proposed. The projection techniques are compared to classic beamforming techniques in term of interference rejection and CSOI recovering. Being usually unknown, the RFI subspace has to be estimated. Several techniques allowing this estimation, based on statistical properties of RFI and cosmic sources (whiteness and cyclostationarity), are also presented and compared. The different techniques have been applied to real astronomical data, provided by the European radio telescope LOFAR. A last section presents an RFI mitigation algorithm implemented on the demonstrator EMBRACE.

La fissure d'engrenages peut être considérée comme un défaut des plus compliqués à diagnostiquer car sa signature vibratoire n'est pas vraiment connue. En plus, l'intégration des fissures dans les modèles numériques n'est pas une tâche simple. D'autre part, le diagnostic des engrenages peut être fait dans le domaine temporel à travers des descripteurs statistiques ou dans le domaine fréquentiel grâce à l'analyse spectrale ou l'analyse cepstrale. Lors de l'apparition d'un défaut de fissure, des phénomènes non linéaires et non-stationnaires se manifestent ce qui rend les outils classiques de traitement du signal moins fiables. Dans ce manuscrit, nous répondons à toutes ces problématiques en développant un modèle d'engrenages à 6 DDL qui porte une fissure qui respire. Le modèle nous permet d'étudier la signature de la fissure et son effet sur les vibrations résultantes indépendamment de l'effet des autres composantes du système. Les résultats ont montré que la fissure conduit à une chute de la rigidité d'engrènement. En plus, la respiration de la fissure cause une fatigue dans le matériau ce qui engendre un terme aléatoire dans le signal vibratoire. La combinaison du terme aléatoire avec la composante périodique due à la rotation des arbres conduit à l'apparition de la cyclostationnarité d'ordre 2. Une étude comparative de sensibilité et de robustesse entre la transformée de Fourier rapide, la cyclostationnarité d'ordre 2 et les estimateurs de la phase instantanée (la transformée de Hilbert, estimation of signal parameters via rotational invariance techniques avec une fenêtre glissante, weighted least squares estimation et le scalogramme de phase) est effectuée pour la détection précoce des fissures. En plus, des essais expérimentaux ont été effectués sur un banc d'essais d'engrenages avec différentes dimensions de fissures. Les résultats théoriques et expérimentaux ont montré que l'analyse cyclostationnaire est la méthode la plus sensible et la plus robuste pour la détection précoce des fissures par rapport aux méthodes proposées. De plus, l'analyse de la phase instantanée donne également des résultats intéressants dans le cas des défauts de fissures. Nous avons montré que le scalogramme de phase est, a priori, plus performant que les autres approches / The gear crack is considered as the most complicated failure to diagnose because its vibration signature is not really known. In addition, the integration of crack defect in numerical models is not a simple task. On the other hand, gears diagnosis can be done in the time domain through statistical descriptors or in the frequency domain using spectral analysis or cepstral analysis. During the appearance of a crack defect, nonlinear and nonstationary phenomena occur which makes the classical tools of signal processing unreliable. In this manuscript, we respond to these challenges by developing a gear model of 6 DOF that has a crack that breathes. This allows us to study the signature of the defect in the resulting vibrations with a flexible way away from external vibrations. The results showed that the crack leads to a fall in the mesh stiffness. In addition, the opening and closing of the crack causes a fatigue in the material which generates a random term in the vibration signal. The combination of the random term with periodic component due to the rotation of the shafts leading to the appearance of second-order cyclostationary. A comparative study of sensitivity and robustness between the fast Fourier transform, second-order cyclostationary and estimators of instantaneous phase (the Hilbert transform, Estimation of Signal Parameters via Rotational Invariance Techniques with a sliding window, Weighted Least Squares Estimation and phase scalogramme) is performed for the early detection of cracks. In addition, experimental tests were carried out on a test-bench with different sizes of crack. The theoretical and experimental results showed that the cyclostationary analysis is the most sensitive and most robust method for the early detection of cracks in comparison with the other evaluated methods. Furthermore, the analysis of the instantaneous phase also gives good results in the case of crack defects. We have shown that the phase scalogramme is a priori more efficient than other approaches

With the recent developments in technology, information and communication technologies (ICTs) are becoming more widespread and one of the basic building blocks of every humans life. The increasing demand in broadband communication calls for new technologies. Power line communication (PLC) is one of the potential candidates for next generation ICTs. Although communication through power lines has been investigated for a long time, PLC systems were never taken into account seriously because of its harsh communication medium. However, with the development of more robust data transmission schemes, communication over the power lines is becoming a strong alternative technology because of the existence of the infrastructure and the ubiquity of the network. In order to establish reliable communication systems operating on power line networks (PLNs), characteristics of power line channels have to be investigated very carefully. Unpredictable characteristics of PLNs seriously affect the performance of communication systems. Similar to the other communication channels, PLC environment is affected by noise, attenuation, and multipath type of channel distortions. The level of noise in PLNs is much higher than any other type of communication networks. Furthermore, the frequency dependent attenuation characteristics of power lines and multipath stemming from impedance mismatches are the other distortion factors which have to be investigated in order to establish a reliable PLC system. In this thesis, we focus on modeling of noise, frequency dependent attenuation, and multipath characteristics of power line channels within the frequency range between 30kHz and 30MHz.

With the rapid and ubiquitous deployment of mobile communications in recent years, cochannel interference has become a critical problem because of its impact on system capacity and quality of service. The conventional approach to minimizing interference is through better cell planning and design. Digital Signal Processing COSP) based interference rejection techniques provide an alternative approach to minimize interference and improve system capacity. Single channel adaptive interference rejection techniques have long been used for enhancing digitally modulated signals. However these techniques are not well suited for analog mobile phone system (AMPS) and narrowband AMPS (NAMPS) signals because of the large spectral overlap of the signals of interest with interfering signals and because of the lack of a well defined signal structure that can be used to separate the signals. Our research has created novel interference rejection techniques based on time-dependent filtering which exploit spectral correlation characteristics exhibited by AMPS and NAMPS signals. A mathematical analysis of the cyclostationary features of AMPS and NAMPS signals is presented to help explain and analyze these techniques. Their performance is investigated using both simulated and digitized data. The impact of these new techniques on AMPS system capacity is also studied. The adaptive algorithms and structures are refined to be robust in various channel environments and to be computationally efficient. / Ph. D.

In this dissertation novel methods are proposed for communicating in interference limited environments as well as detecting such interference. The methods include introducing redundancies into multicarrier signals to make them more robust, applying a novel filtering structure for mitigating radar interference to orthogonal frequency division multiplexing (OFDM) signals and for exploiting the cyclostationary nature of signals to whiten the spectrum in blind signal detection. Data symbols are repeated in both time and frequency across orthogonal frequency division multiplexing (OFDM) symbols, creating a cyclostationary nature in the signal. A Frequency Shift (FRESH) filter can then be applied to the cyclostationary signal, which is the optimal filter and is able to reject interference much better than a time-invariant filter such as the Wiener filter. A novel time-varying FRESH filter (TV-FRESH) filter is developed and its Minimum Mean Squared Error (MMSE) filter weights are found. The repetition of data symbols and their optimal combining with the TV-FRESH filter creates an effect of improving the Bit Error Rate (BER) at the receiver, similar to an error correcting code. The important distinction for the paramorphic method is that it is designed to operate within cyclostationary interference, and simulation results show that the symbol repetition can outperform other error correcting codes. Simulated annealing is used to optimize the signaling parameters, and results show that a balance between the symbol repetition and error correcting codes produces a better BER for the same spectral efficiency than what either method could have achieved alone. The TV-FRESH filter is applied to a pulsed chirp radar signal, demonstrating a new tool to use in radar and OFDM co-existence. The TV-FRESH filter applies a set of filter weights in a periodically time-varying fashion. The traditional FRESH filter is periodically time-varying due to the periodicities of the frequency shifters, but applies time-invariant filters after optimally combine any spectral redundancies in the signal. The time segmentation of the TV-FRESH filter allows spectral redundancies of the radar signal to be exploited across time due to its deterministic nature. The TV-FRESH filter improves the rejection of the radar signal as compared to the traditional FRESH filter under the simulation scenarios, improving the SINR and BER at the output of the filter. The improvement in performance comes at the cost of additional filtering complexity. A time-varying whitening filter is applied to blindly detect interference which overlaps with the desired signal in frequency. Where a time-invariant whitening filter shapes the output spectrum based on the power levels, the proposed time-varying whitener whitens the output spectrum based on the spectral redundancy in the desired signal. This allows signals which do not share the same cyclostationary properties to pass through the filter, improving the sensitivity of the algorithm and producing higher detection rates for the same probability of false alarm as compared to the time-invariant whitener. / Ph. D. / This dissertation proposes novel methods for building robust wireless communication links which can be used to improve their reliability and resilience while under interference. Wireless interference comes from many sources, including other wireless transmitters in the area or devices which emit electromagnetic waves such as microwaves. Interference reduces the quality of a wireless link and depending on the type and severity may make it impossible to reliably receive information. The contributions are both for communicating under interference and being able to detect interference. A novel method for increasing the redundancy in a wireless link is proposed which improves the resiliency of a wireless link. By transmitting additional copies of the desired information the wireless receiver is able to better estimate the original transmitted signal. The digital receiver structure is proposed to optimally combine the redundant information, and simulation results are used to show its improvement over other analogous methods. The second contribution applies a novel digital filter for mitigating interference from a radar signal to an Orthogonal Frequency Division Multiplexing (OFDM) signal, similar to the one which is being used in Long Term Evolution (LTE) mobile phones. Simulation results show that the proposed method out performs other digital filters at the most of additional complexity. The third contribution applies a digital filter and trains it such that the output of the filter can be used to detect the presence of interference. An algorithm which detects interference can tip off an appropriate response, and as such is important to reliable wireless communications. Simulation results are used to show that the proposed method produces a higher probability of detection while reducing the false alarm rate as compared to a similar digital filter trained to produce the same effect.

The advent of spectrum sharing has increased the need for robust interference rejection methods. The Citizens Broadband Radio Service (CBRS) band is soon to be occupied by LTE waveforms and License Assisted Access (LAA) will have LTE signals coexisting with other signals in the 5 GHz band. In anticipation of this need, we present a method for interference rejection of cyclostationary signals, which can also help avoid interference through better detection of low power co-channel signals. The method proposed in this thesis consists of a frequency-shift (FRESH) filter which acts as a whitening filter, canceling the interference by exploiting its cyclostationarity. It learns the cyclostationary characteristics of the interferer blindly, through a property restoration algorithm which aims to drive the spectrum to white noise. The property restoration algorithm, inspired by the constant modulus algorithm (CMA), is applied to each frequency bin to determine the optimal coefficients for the proposed CMA FRESH whitening filter (CFW). The performance of the CFW in interference rejection is compared to a time-invariant version, and proposed use cases are analyzed. The use cases consist of the rejection of a high powered, wider bandwidth interferer which is masking the signal-of-interest (SOI). The interferer is rejected blindly, with no knowledge of its characteristics. We analyzed signal detection performance in the case that the SOI is another user with much lower power, for multiple types of SOIs ranging from BPSK to OFDM. We also deal with the case that the SOI is to be received and demodulated; we recover it and compare resulting bit error rates to state of the art FRESH filters. The results show significantly better signal detection and recovery. / Master of Science / Wireless communication is complicated by the fact that multiple radios may be attempting to transmit at the same frequency, time and location concurrently. This scenario may be a due to malicious intent by certain radios (jamming), or mere confusion due to a lack of knowledge that another radio is transmitting in the same channel. The latter scenario is more common due to congested wireless spectrum, as the number of devices increases exponentially. In either case, interference results. We present a novel interference rejection method in this work, one that is blind to the properties of the interferer and adapts to cancel it. It follows the philosophy of property restoration as extolled by the constant modulus algorithm (CMA) and is a frequency shift (FRESH) filter, hence the name. The process of restoring the wireless spectrum to white noise is what makes it a whitening filter, and is also how it adapts to cancel interference. Such a filter has myriad possible uses, and we examine the use case of rejecting interference to detect or recover the signal-of-interest (SOI) that we are attempting to receive. We present performance results in both cases and compare with conventional time-invariant filters and state of the art FRESH filters.

This research investigates signal processing involving a single electromagnetic vector-sensor, with an emphasis on the problem regarding signal-selective narrowband direction-of-arrival (DOA) estimation in the presence of interference. The approach in this thesis relies on a high-resolution ESPRIT-based algorithm. Unlike spatially displaced arrays, the sensor cannot estimate the DOA of sources using phase differences between the array elements, as the elements are spatially co-located. However, the sensor measures the full electromagnetic field vectors, so the DOA can be estimated through the Poynting vector. Limited information is available in the open literature regarding signal-selective DOA estimation for a single electromagnetic vector-sensor. In this thesis, it is shown how the Uni-Vector-Sensor-ESPRIT (UVS-ESPRIT) algorithm that relies on a time-series invariance and was originally devised for deterministic harmonic sources can be applied to non-deterministic sources. Additionally, two algorithms, one based on cyclostationarity and the other based on fourth-order cumulants, are formulated based on the UVS-ESPRIT algorithm and are capable of selectively estimating the source DOA in the presence of interference based on the statistical properties of the sources. The cyclostationarity-based UVS-ESPRIT algorithm is capable of selectively estimating the signal-of-interest DOA when the sources have the same carrier frequency, and thus overlap in frequency. The cumulant-based UVS-ESPRIT algorithm devised for this sensor relies on the independent component analysis algorithm JADE and is capable of selectively estimating the signal-of-interest DOA through the fourth-order cumulants only, is robust to spatially colored noise, and is capable of estimating the DOA of more sources than sensor elements. / Master of Science / Electromagnetic vector-sensors are specialized sensors capable of capturing the full electromagnetic field vectors at a single point in space. Direction-of-arrival (DOA) estimation is the problem of estimating the spatial-angular parameters of one or more wavefronts impinging on an array. For a single electromagnetic vector-sensor, the array elements are not spatially displaced, but it is still possible to estimate the direction-of-arrival through the Poynting vector, which relates the electric and magnetic field vectors to the direction of propagation of an electromagnetic wave. Although direction-of-arrival estimation is a well-established area of research, there is limited discussion in the open literature regarding signal-selective DOA estimation in the presence of interference for a single electromagnetic vector-sensor. This research investigates this problem and discusses how the high-resolution Uni-Vector-Sensor-ESPRIT (UVS-ESPRIT) algorithm may be applied to non-deterministic sources. ESPRIT based algorithms capable of selectively estimating the source DOA are formulated based on the cyclostationarity and higher-order statistics of the sources, which are approaches known to be robust to interference. The approach based on higher-order statistics is also robust to spatially colored noise and is capable of estimating the DOA of more sources than sensor elements. The formulation of the UVS-ESPRIT for higher-order statistics relies on the application of the independent component analysis algorithm JADE, an unsupervised learning technique. Overall, this research investigates signal-selective direction-of-arrival estimation using an ESPRIT-based algorithm for a single electromagnetic vector-sensor.

Les travaux présentés dans cette mémoire visent à développer de nouvelles méthodes qui exploitent les propriétés de cyclostationnarité pour traiter des signaux de force de réaction du sol enregistrées au cours de la marche et la course à pied. Nous nous intéressons à l’analyse de la locomotion humaine dans trois domaines d´études: une étude liée à la pathologie, une deuxième liée directement à l’âge et une troisième relative à la fatigue. En effet, la détection du risque de chute chez les personnes âgées pour fin de prévention contre la chute constitue un enjeu majeur, car cette chute entraine d’une part un nombre de décès important et d’autres part se traduit par un cout élevée de la santé publique. Par ailleurs, l’étude de la fatigue musculaire en particulier pour l’amélioration des performances des sportifs de haut niveau a fait l’objet de nombreux travaux de recherche & développement. La recherche et le développement de nouvelles méthodes et d’indicateurs dans le domaine de traitement de signal dans le but de caractériser la locomotive humaine, permettrait des avancées intéressantes dans les enjeux évoqués ci-dessus. La complexité des signaux GRF est définie par le système neuromusculaire qui génère ce signal. Une meilleure connaissance de ce système nécessite le développement des méthodes de séparation de sources et des outils avancés de traitement du signal pour mieux décrire le système considéré. En effet, nous montrons dans cette thèse que les signaux GRF peuvent être modélisés dans un cadre cyclostationnaire élargi. Les composantes de signal GRF (contribution active et passive) sont séparées par de nouvelles techniques de séparation de sources. Cette modélisation ouvre de nouvelles perspectives pour la décomposition et identification des sources individuelles. D'autre part, on exploite les caractères cyclostationnaire des signaux dans le cadre de la méthode d'analyse en composantes morphologique (MCA). Cet algorithme nous permet de séparer avec succès les composantes d’ordre 1 et d’ordre 2 des signaux considérés. Finalement, nous nous proposons un nouveau modèle utile pour l'étude et la caractérisation de cyclostationnarité. Il présente l'effet de la variation aléatoire de la pente sur le spectre du signal cyclique. Nous appelons ce modèle (modèle cyclostationnaire à pente aléatoire). Nous appliquons ce modèle pour l'étude des signaux biomécaniques où nous considérons la pente comme une mesure spécifique extraite des forces de réaction du sol. Les résultats montrent que la pente et les polynômes à coefficients aléatoires du pic passive peuvent jouer un rôle important et fournir des informations intéressantes concernant la fatigue et concernant la performance de marche et course à pied / The research work presented in this dissertation, involves the development of novel methodologies and methods, for the exploitation of cyclostationarity properties and for the treatment of ground reaction force signals, recorded during walking and running. We are especially interested in the analysis of human locomotion in three fields of interest: a study relating to pathology, a study directly related to age, and a study of muscle fatigue. Indeed, the detection of risk of falling among the elderly for the prevention of falls is of major concern. This is because falling on the one hand leads to a large number of deaths and secondly, resulting in higher costs of public health.Study the muscle fatigue in particular has occupied taken a big share out of this research due to the importance of such events like strenuous level of sports. Research and development of new methods and indicators in the field of signal processing for better characterizing the human locomotion, would allow interesting advances in the aforementioned issues. The complexity of GRF signals is defined by the neuromuscular system which generates this signal. Improved knowledge of this system requires developing source separation methods and advanced signal processing tools to better describe the system under consideration. Indeed, we will endeavor to show in this dissertation that GRF signals can be modeled within an enlarged cyclostationary framework. The GRF signal components (active and passive contribution) are separated by means of new source separation techniques. This modeling opens new perspectives for the decomposition and identification of individual sources. On the other hand, we exploit the cyclostationary characters of signals in the context of Morphological component analysis (MCA) method. Such algorithm enables us to successfully separate the first and second order components of the signals under consideration. Finally, we provide a new model useful for studying and characterizing cyclostationarity. It presents the impact of random slope variation on the cyclic spectrum of the signal. We call this model the random slope modulation (RSM). We apply this model for studying biomechanical signals where we consider the slope as a specic measure extracted from the vertical ground reaction forces. The results show that the slope and polynomial random coefficients of passive peaks can play important role and provide interesting information concerning fatigue and concerning running / walking performance

Les signaux multicanaux sont des signaux captés à travers plusieurs canaux ou capteurs, portant chacun un mélange de sources, une partie desquelles est connue alors que le reste des sources reste inconnu. Les méthodes à l’aide desquelles l’isolement ou la séparation des sources est accomplie sont connues par les méthodes de séparation de sources en général, et si le degré d’inconnu est large, par la séparation aveugle des sources (SAS). Cependant, la SAS appliquée aux signaux multicanaux est en fait plus facile de point de vue mathématique que l’application de la SAS sur des signaux monocanaux, ou un seul capteur existe et tous les signaux arrivent au même point pour enfin produire un mélange de sources inconnues. Tel est le domaine de cette thèse. Nous avons développé une nouvelle technique de SAS : une combinaison de plusieurs méthodes de séparation et d’optimisation, basée sur la factorisation non-négative des matrices (NMF). Cette méthode peut être utilisée dans de nombreux domaines comme l’analyse des sons et de la parole, les variations de la bourse, et les séismographes. Néanmoins, ici, les signaux de force de réaction de terre verticaux (VGRF) monocanaux d’un groupe d’athlètes coureurs d’ultra-marathon sont analysés et séparés pour l’extraction du peak passif du peak actif d’une nouvelle manière adaptée à la nature de ces signaux. Les signaux VGRF sont des signaux cyclo-stationnaires caractérisés par des double-peaks, chacun étant très rapide et parcimonieux, indiquant les phases de course de l’athlète. L’analyse des peaks est extrêmement importante pour déterminer et prédire la condition du coureur : problème physiologique, problème anatomique, fatigue etc. De plus, un grand nombre de chercheurs ont prouvé que l’impact du pied postérieur avec la terre d’une manière brutale, l’analyse de ce phénomène peut nous ramener à une prédiction de blessure interne. Ils essayent même d’adopter une technique de course - Non-Heel-strike Running (NHS) - par laquelle ils obligent les coureurs à courir sur le pied-antérieur seulement. Afin d'étudier ce phénomène, la séparation du peak d’impact du VGRF permet d'isoler la source portant les informations patho-physiologiques et le degré de fatigue. Nous avons introduit de nouvelles méthodes de prétraitement et de traitement des signaux VGRF pour remplacer le filtrage de bruit traditionnel utilisé partout, et qui peut parfois détruire les peaks d’impact qui sont nos sources à séparer, base sur le concept de soustraction spectrale pour le filtrage, utilisée avec les signaux de parole, après l’application d’un algorithme d’échantillonnage intelligent et adaptatif qui décompose les signaux en pas isolés. Une analyse des signaux VGRF en fonction du temps a été faite pour la détection et la quantification de la fatigue des coureurs durant les 24 heures de course. Cette analyse a été accomplie au domaine fréquentiel/spectral où nous avons détecté un décalage clair du contenu fréquentiel avec la progression de la course indiquant la progression de la fatigue. Nous avons défini les signaux cyclosparse au domaine temporel, puis traduit cette définition à son équivalent au domaine temps-fréquence utilisant la transformée Fourier a court-temps (STFT). Cette représentation a été décomposée à travers une nouvelle méthode que l’on a appelé Cyclosparse Non-negative Matrix Factorisation (Cyclosparse-NMF), basée sur l’optimisation de la minimisation de la divergence Kullback-Leibler (KL) avec pénalisation liée à la périodicité et la parcimonie des sources, ayant comme but final d’extraire les sources cyclosparse du mélange monocanal appliquée aux signaux VGRF monocanaux. La méthode a été testée sur des signaux analytiques afin de prouver l’efficacité de l’algorithme. Les résultats se sont avéré satisfaisants, et le peak impact a été séparé du mélange VGRF monocanal. / The purpose of the presented work is to develop a customized Single-channel Blind Source Separation technique that aims to separate cyclostationary and transient pulse-like patterns/sources from a linear instantaneous mixture of unknown sources. For that endeavor, synthetic signals of the mentioned characteristic were created to confirm the separation success, in addition to real life signals acquired throughout an experiment in which experienced athletes were asked to participate in a 24-hour ultra-marathon in a lab environment on an instrumented treadmill through which their VGRF, which carries a cyclosparse Impact Peak, is continuously recorded with very short discontinuities during which blood is drawn for in-run testing, short enough not to provide rest to the athletes. The synthetic and VGRF signals were then pre-processed, processed for Impact Pattern extraction via a customized Single-channel Blind Source Separation technique that we termed Cyclo-sparse Non-negative Matrix Factorization and analyzed for fatigue assessment. As a result, the Impact Patterns for all of the participating athletes were extracted at 10 different time intervals indicating the progression of the ultra-marathon for 24 hours, and further analysis and comparison of the resulting signals proved major significance in the field of fatigue assessment; the Impact Pattern power monotonically increased for 90% of the subjects by an average of 24.4 15% with the progression of the ultra-marathon during the 24-hour period. Upon computation of the Impact Pattern separation algorithm, fatigue progression showed to be manifested by an increase in reliance on heel-strike impact to push to the bodyweight as a compensation for the decrease in muscle power during propulsion at toe-off. This study among other presented work in the field of VGRF processing forms methods that could be implemented in wearable devices to assess and track runners’ gait as a part of sports performance analysis, rehabilitation phase tracking and classification of healthy vs. unhealthy gait.

'It is not a lack of spectrum. It is an issue of efficient use of the available spectrum"--conclusions of the FCC Spectrum Policy Task Force. There is growing interest towards providing broadband communication with high bit rates and throughput, especially in the ISM band, as it was an ignition of innovation triggered by the FCC to provide, to some extent, a regulation-free band that anyone can use. But with such freedom comes the risk of interference and more responsibility to avoid causing it. Therefore, the need for accurate interference detection and identification, along with good blind detection capabilities are inevitable. Since cognitive radio is being adopted widely as more researchers consider it the ultimate solution for efficient spectrum sharing [1], it is reasonable to study the cognitive radio in the ISM band [2]. Many indications show that the ISM band will have less regulation in the future, and some even predict that the ISM may be completely regulation free [3]. In the dawn of cognitive radio, more knowledge about possible interfering signals should play a major role in determining optimal transmitter configurations. Since signal identification and interference will be the core concerns [4], [5], we will describe a novel approach for a cognitive radio spectrum sensing engine, which will be essential to design more efficient ISM band transceivers. In this thesis we propose a novel spectrum awareness engine to be integrated in the cognitive radios. Furthermore, the proposed engine is specialized for the ISM band, assuming that it can be one of the most challenging bands due to its free-to-use approach. It is shown that characterization of the interfering signals will help with overcoming their effects. This knowledge is invaluable to help choose the best configuration for the transceivers and will help to support the efforts of the coexistence attempts between wireless devices in such bands.