In a multiuser DS-CDMA system with frequency selectivity, each user’s spreading sequence is transmitted through a different channel and the autocorrelation and the cross correlation properties of the received sequences will not be the same as that of the transmitted sequences. The best way of designing spreading sequences for frequency selective channels is to design them at the receiver exploiting the users’ channel characteristics. By doing so, we can show that the designed sequences outperform single user AWGN performance. In existing sequence design algorithms for frequency selective channels, the design is done in the time domain and the connection to frequency domain properties is not established. We approach the design of spreading sequences based on their frequency domain characteristics. Based on the frequency domain characteristics of the spreading sequences with unconstrained amplitudes and phases, we propose a reduced-rank sequence design algorithm that reduces the computational complexity, feedback bandwidth and improves the performance of some existing sequence design algorithms proposed for frequency selective channels. We propose several different approaches to design the spreading sequences with constrained amplitudes and phases for frequency selective channels. First, we use the frequency domain characteristics of the unconstrained spreading sequences to find a set of constrained amplitude sequences for a given set of channels. This is done either by carefully assigning an already existing set of sequences for a given set of users or by mapping unconstrained sequences onto a unit circle. Secondly, we use an information theoretic approach to design the spreading sequences by matching the spectrum of each user’s sequence to the water-filling spectrum of the user’s channel. Finally, the design of inner shaping codes for single-head and multi-head magnetic recoding channels is discussed. The shaping sequences are designed considering them as short spreading codes matched to the recoding channels. The outer channel code is matched to the inner shaping code using the extrinsic information transfer chart analysis. In this dissertation we introduce a new frequency domain approach to design spreading sequences for frequency selective channels. We also extend this proposed technique to design inner shaping codes for partial response channels.

We present a serially concatenated coding scheme for partial response channels. The encoder consists of an outer irregular LDPC code and an inner matched spectrum trellis code. These codes are shown to offer considerable improvement over the i.i.d. capacity (> 1 dB) of the channel for low rates (approximately 0.1 bits per channel use). We also present a qualitative argument on the optimality of these codes for low rates. We also formulate a performance index for such codes to predict their performance for low rates. The results have been verified via simulations for the (1-D)/sqrt(2) and the (1-D+0.8D^2)/sqrt(2.64) channels. The structure of the encoding/decoding scheme is considerably simpler than the existing scheme to maximize the information rate of encoders over partial response channels.

We present a serially concatenated coding scheme for partial response channels. The encoder consists of an outer irregular LDPC code and an inner matched spectrum trellis code. These codes are shown to offer considerable improvement over the i.i.d. capacity (> 1 dB) of the channel for low rates (approximately 0.1 bits per channel use). We also present a qualitative argument on the optimality of these codes for low rates. We also formulate a performance index for such codes to predict their performance for low rates. The results have been verified via simulations for the (1-D)/sqrt(2) and the (1-D+0.8D^2)/sqrt(2.64) channels. The structure of the encoding/decoding scheme is considerably simpler than the existing scheme to maximize the information rate of encoders over partial response channels.

Tandem mass spectrometry, also known as MS/MS, is an analytical technique to measure the mass-to-charge ratio of charged ions and widely used in genomics, proteomics and metabolomics areas. There are two types of automatic ways to interpret tandem mass spectra: de novo methods and database searching methods. Both of them need to use massive computational resources and complicated comparison algorithms. The real-time peptide-spectrum matching (RT-PSM) algorithm is a database searching method to interpret tandem mass spectra with strict time constraints. Restricted by the hardware and architecture of an individual workstation the RT-PSM algorithm has to sacrifice the level of accuracy in order to provide prerequisite processing speed. The peptide-spectrum similarity scoring module is the most time-consuming part out of four modules in the RT-PSM algorithm, which is also the core of the algorithm. In this study, a multi-core computing algorithm is developed for individual workstations. Moreover, a distributed computing algorithm is designed for a cluster. The improved algorithms can achieve the speed requirement of RT-PSM without sacrificing the accuracy. With some expansion, this distributed computing algorithm can also support different PSM algorithms. Simulation results show that compared with the original RT-PSM, the parallelization version achieves 25 to 34 times speed-up based on different individual workstations. A cluster with 240 CPU cores could accelerate the similarity score module 210 times compare with the single-thread similarity score module and the whole peptide identification process 85 times compare with the single-thread peptide identification process.

Les signaux accélérométriques enregistrés lors de l’occurrence d’un événement sismique est très large présentent une forte variabilité, par conséquent ils ne sont pas utilisé dans les analyse dynamiques de tenue sismique des structures. En effet, l’utilisation des accélérogrammes réels, pour les analyses dynamiques non linéaires, s’avère couteuse en termes de temps de calcul. La pratique courante prévoit la minimisation (voir suppression) de telle variabilité, mais les conséquences d’une telle opération sur la réponse des structures ne sont pas clairement indiquées. L’étude ci-présente a pour scope la quantification de l’impact des méthodes de sélection qui gardent la variabilité du signal sur les résultats de l’analyse de la réponse des structures (exprimée en termes d’engineering demand parameters EDPs). En particulier les questions suivantes seront investiguées :Quel est le niveau de variabilité des accélérogrammes réels et comment ce niveau est modifié par les techniques couramment utilisées ?Quelle est l’impact de la variabilité sur la réponse de plusieurs types de structures ?Pour un scénario sismique donné, un spectre cible est défini à partir de plusieurs équation de prédiction du mouvement sismique, sélection parmi celles disponibles en littérature. Les accélérogrammes sont sélectionnés à partir de quatre familles d’accélérogrammes, chacune relative à une méthode de modification : réels (enregistrés); mise à l’échelle (multiplication, par un facteur) ; calés aux spectres cibles avec large tolérance ; calés aux spectres cibles dans une plage de tolérance étroite.Chaque jeu de signaux est composé de cinq accélérogrammes et la sélection des signaux est faite en tenant compte de deux sources de variabilité : la variabilité au sein de chaque jeu de données (intraset), et la variabilité entre les différents jeux de données (interset) tous compatibles avec le même spectre cible. Les tests sur les EDPs menés sur les signaux accélérométriques réels mènent à la quantification de la variabilité naturelle (pour le scénario considéré). Les analyses basées sur les signaux réels sont utilisés comme benchmark afin d’évaluer non seulement de combien la distribution des EDPs (en termes de valeur moyenne et variabilité) est réduite par les différentes méthodes testées, mais aussi d’évaluer l’impact des choix de l’équation de prédiction du mouvement, des plages de tolérance, du nombre d’accélérogrammes constituant chaque jeu, du nombre de jeux, de le scope de l’analyse structurale et le modèle de structure.Ce travaille nous conduit à conclure que un seul jeu d’accélérogramme, tel qu’utilisé dans la pratique courante, est insuffisant pour assurer le niveau d’EDPs indépendamment de la méthode de modification utilisés, cela est lié à la variabilité des signaux et entre les jeux d’accélérogrammes. Les signaux réels, compatibles avec le spectre définis pour le scénario sismique, are l’option plus réaliste pour l’analyse dynamique non-linéaire ; si une méthode de modification du signal est nécessaire, la plus adaptées dépend du scope de l’analyse spectrale et du modèle. Le choix de l’équation de prédiction du mouvement sismique utilisée pour définir le spectre cible impacte significativement les caractéristiques des mouvements sismiques et des EDPs. Cette observation ne dépend pas de la stratégie de de modification du signal. / The observed variability is very large among natural earthquake records, which are not consolidated in the engineering applications due to the cost and the duration. In the current practice with the nonlinear dynamic analysis, the input variability is minimized, yet without clear indications of its consequences on the output seismic behavior of structures. The study, herein, aims at quantifying the impact of ground motion selection with large variability on the distribution of engineering demand parameters (EDPs) by investigating the following questions:What is the level of variability in natural and modified ground motions?What is the impact of input variability on the EDPs of various structural types?For a given earthquake scenario, target spectra are defined by ground motion prediction equations (GMPEs). Four ground motion modification and selection methods such as (1) the unscaled earthquake records, (2) the linearly scaled real records, (3) the loosely matched spectrum waveforms, and (4) the tightly matched waveforms are utilized. The tests on the EDPs are performed on a record basis to quantify the natural variability in unscaled earthquake records and the relative changes triggered by the ground motion modifications.Each dataset is composed by five accelerograms; the response spectrum compatible selection is then performed by considering the impact of set variability. The intraset variability relates to the spectral amplitude dispersion in a given set, and the interset variability relates to the existence of multiple sets compatible with the target.The tests on the EDPs are performed on a record basis to quantify the natural variability in unscaled earthquake records and the relative changes triggered by the ground motion modifications. The distributions of EDPs obtained by the modified ground motions are compared to the observed distribution by the unscaled earthquake records as a function of ground motion prediction equations, objective of structural analysis, and structural models.This thesis demonstrates that a single ground motion set, commonly used in the practice, is not sufficient to obtain an assuring level of the EDPs regardless of the GMSM methods, which is due to the record and set variability. The unscaled real records compatible with the scenario are discussed to be the most realistic option to use in the nonlinear dynamic analyses, and the ‘best’ ground motion modification method is demonstrated to be based on the EDP, the objective of the seismic analysis, and the structural model. It is pointed out that the choice of a GMPE can provoke significant differences in the ground motion characteristics and the EDPs, and it can overshadow the differences in the EDPs obtained by the GMSM methods.

No description available.