Compressed sensing is an emerging field, which proposes that a small collection of linear projections of a sparse signal contains enough information for perfect reconstruction of the signal. In this thesis, we study the general problem of modeling and reconstructing spatially or temporally correlated sparse signals in a distributed scenario. The correlation among signals provides an additional information, which could be captured by joint sparsity models. After modeling the correlation, we propose two different reconstruction algorithms that are able to successfully exploit this additional information. The first algorithm is a very fast greedy algorithm, which is suitable for large scale problems and can exploit spatial correlation. The second algorithm is based on a thresholding algorithm and can exploit both the temporal and spatial correlation. We also generalize the standard joint sparsity model and propose a new model for capturing the correlation in the sensor networks.

Static Single Assignment (SSA) is an Intermediate Representation (IR) that simplifies the design and implementation of analyses and optimizations. While intraprocedural SSA is ubiquitous in modern compilers, the use of interprocedural SSA (ISSA), although seemingly a natural extension, is limited. In this dissertation, we propose new techniques to construct and integrate ISSA into modern compilers and evaluate the benefit of using ISSA form. First, we present an algorithm that converts the IR into ISSA form by introducing new instructions. To our knowledge, this is the first IR-based ISSA proposed in the literature. Moreover, in comparison to previous work we increase the number of SSA variables, extend the scope of definitions to the whole program, and perform interprocedural copy propagation. Next, we propose an out-of-ISSA translation that simplifies the integration of ISSA form into a compiler. Our out-of-ISSA translation algorithm enables us to leverage ISSA to improve performance without having to update every compiler pass. Moreover, we demonstrate the benefit of ISSA for a number of compiler optimizations. Finally, we present an ISSA-based interprocedural induction variable analysis. Our implementation introduces only a few changes to the SSA-based implementation while enabling us to identify considerably more induction variables and compute more loop trip counts.

One goal of this thesis is to address several challenging compact antenna design issues by using transmission-line metamaterials. In particular, we demonstrate the design of a compact antenna with an extended bandwidth, multiband/multifunction compact/small antennas, and mutual-coupling reduction for two closely-spaced small antennas. The proposed compact transmission-line metamaterial antenna employs the concept of zeroth- index resonance and a wideband characteristic is enabled by detuning the resonance of each constituent metamaterial unit cell at a slightly different frequency, thus creating a multi-resonant return-loss passband. Furthermore, a single-cell transmission-line metamaterial loading scheme is applied to regular printed monopole antennas in order to introduce additional resonances at the low band and create multiband small antennas that meet the specifications for WiFi and WiMAX applications. Lastly, a simple ap- proach for reducing the mutual coupling in two closely-spaced small antennas is also presented, based on the idea of self-cancelation of the induced currents. The other important goal of this thesis is to develop volumetric negative-refractive- index transmission-line (NRI-TL) metamaterials. A volumetric NRI-TL slab is created by stacking 2D NRI transmission-line grids in the shunt-node configuration. This is done in a simple manner through images induced in a parallel-plate environment. Additional vias are strategically placed to suppress the parasitic parallel-plate mode. Moreover, multiconductor transmission-line theory is used to model the volumetric metamaterial slab. A fully-printed volumetric Veselago-Pendry transmission-line lens is designed and matched to free space. Using this proposed lens, it has been experimentally verified that the diffraction limit can be overcome.

This thesis presents a new family of converters for high power interconnection of dc buses with different voltage levels. Proposed converters achieve high voltage dc-dc conversion without an intermediate ac conversion stage. This function is implemented without series connection of active switches, or the use of isolation transformers. The salient features of proposed converters are (i) design and construction simplicity, (ii) low switching losses through soft turn-on and soft turn-off, (iii) single stage dc-dc conversion without high-current chopping, (iv) modular structure, (v) equal voltage sharing among the converter modules. Three converter circuits are investigated. The first performs unidirectional power transfer from a dc bus with higher voltage to a dc bus with lower voltage. The second performs unidirectional power transfer from a dc bus with lower voltage to a dc bus with higher voltage. Both converters are suitable for interconnecting single pole dc buses with same polarity, or double pole dc buses. A third converter is also presented which performs the function of either the first or the second converter with polarity reversal. The third converter is suitable for interconnecting single pole dc buses with different polarities, or double pole dc buses. By hybrid integration of the proposed three converters, the thesis also investigates other topologies for bidirectional power transfer between two dc buses. Proposed converters operate only in discontinuous conduction mode and exhibit soft switching operation for the active and passive switches. A common feature between the proposed converters is the self current turn-off for the active switches at zero voltage. This allows the use of thyristors as active switches alleviating their reverse recovery losses. For each converter topology, the structure is presented, its operation principle is explained and a complete set of design equations are derived. Comparisons are performed on high-power and high-voltage design examples. The merits and limitations of each converter are concluded. Practical considerations regarding components selection, loss analysis, filter design and the non-idealities of the circuits are studied. Experimental implementation of scaled-down laboratory prototypes is presented to provide a proof of concept and validate the operation principle of the proposed converter topologies.

As biometric recognition becomes increasingly popular, the fear of circumvention, obfuscation and replay attacks is a rising concern. Traditional biometric modalities such as the face, the fingerprint or the iris are vulnerable to such attacks, which defeats the purpose of biometric recognition, namely to employ physiological characteristics for secure identity recognition. This thesis advocates the use the electrocardiogram (ECG) signal for human identity recognition. The ECG is a vital signal of the human body, and as such, it naturally provides liveness detection, robustness to attacks, universality and permanence. In addition, ECG inherently satisfies uniqueness requirements, because the morphology of the signal is highly dependent on the particular anatomical and geometrical characteristics of the myocardium in the heart. However, the ECG is a continuous signal, and this presents a great challenge to biometric recognition. With this modality, instantaneous variability is expected even within recordings of the same individual due to a variety of factors, including recording noise, or physical and psychological activity. While the noise and heart rate variations due to physical exercise can be addressed with appropriate feature extraction, the effects of emotional activity on the ECG signal are more obscure. This thesis deals with this problem from an affective computing point of view. First, the psychological conditions that affect the ECG and endanger biometric accuracy are identified. Experimental setups that are targeted to provoke active and passive arousal as well as positive and negative valence are presented. The empirical mode decomposition (EMD) is used as the basis for the detection of emotional patterns, after adapting the algorithm to the particular needs of the ECG signal. Instantaneous frequency and oscillation features are used for state classification in various clustering setups. The result of this analysis is the designation of psychological states which affect the ECG signal to an extent that biometric matching may not be feasible. An updating methodology is proposed to address this problem, wherein the signal is monitored for instantaneous changes that require the design of a new template. Furthermore, this thesis presents the enhanced Autocorrelation- Linear Discriminant Analysis (AC/LDA) algorithm for feature extraction, which incorporates a signal quality assessment module based on the periodicity transform. Three deployment scenarios are considered namely a) small-scale recognition systems, b) large-scale recognition systems and c) recognition in distributed systems. The enhanced AC/LDA algorithm is adapted to each setting, and the advantages and disadvantages of each scenario are discussed. Overall, this thesis attempts to provide the necessary algorithmic and practical framework for the real-life deployment of the ECG signal in biometric recognition.

This thesis presents network design and operations algorithms suitable for use in an autonomic management system for communication networks with emphasis on network robustness. We model a communication network as a weighted graph and we use graph-theoretical metrics such as network criticality and algebraic connectivity to quantify robustness. The management system under consideration is composed of slow and fast control loops, where slow loops manage slow-changing issues of the network and fast loops react to the events or demands that need quick response. Both of control loops drive the process of network management towards the most robust state. We fist examine the topology design of networks. We compare designs obtained using different graph metrics. We consider well-known topology classes including structured and complex networks, and we provide guidelines on the design and simplification of network structures. We also compare robustness properties of several data center topologies. Next, the Robust Survivable Routing (RSR) algorithm is presented to assign working and backup paths to online demands. RSR guarantees 100% single-link-failure recovery as a path-based survivable routing method. RSR quanti es each path with a value that represents its sensitivity to incremental changes in external traffic and topology by evaluating the variations in network criticality of the network. The path with best robustness (path that causes minimum change in total network criticality) is chosen as primary (secondary) path. In the last part of this thesis, we consider the design of robust networks with emphasis on minimizing vulnerability to single node and link failures. Our focus in this part is to study the behavior of a communication network in the presence of node/link failures, and to optimize the network to maximize performance in the presence of failures. For this purpose, we propose new vulnerability metrics based on the worst case or the expected value of network criticality or algebraic connectivity when a single node/link failure happens. We show that these vulnerability metrics are convex (or concave) functions of link weights and we propose convex optimization problems to optimize each vulnerability metric. In particular, we convert the optimization problems to SDP formulation which leads to a faster implementation for large networks.

Future generation of wireless systems will feature high data rates and be implemented in low voltage CMOS technologies. Direct conversion receivers (DCRs) will be used in such systems which will require low voltage RF front-ends with adequate linearity. The down-converter in a DCR is a critical block in determining linearity. In addition to detailed DCR modeling in MATLAB, this thesis, completed in 2005, deals with the design and characterization of a 1V, 8GHz quadrature down-converter. It consists of two mixers and a quadrature generator implemented in a 0.18m CMOS technology. The mixer architecture proposed in this work uses a new trans-conductor. It simultaneously satisfies the low voltage and high linearity requirements. It also relaxes the inherent trade-off between gain and linearity governing CMOS active mixers. The implemented mixer occupies an area of 320 x 400 m2 and exhibits a power conversion gain of +6.5dB, a P-1dB of -5.5dBm, an IIP3 of +3.5dBm, an IIP2 of better than +48dBm, a noise figure of 11.5dB, an LO to RF isolation of 60dB at 8GHz and consumes 6.9mW of power from a 1V supply. The proposed quadrature generator circuit features a new architecture which embeds the quadrature generation scheme into the LO-buffer using active inductors. The circuit offers easy tune-ability for process, supply and temperature variations by relaxing the coupling between amplitude and phase tuning of the outputs. The implemented circuit occupies an area of 150 x 90m2 and exhibits an amplitude and quadrature phase accuracy of 1 dB and 1.5° respectively over a bandwidth of 100 MHz with a power consumption of 12mW from a 1V supply including the LO-buffer. The quadrature down-converter features an image rejection ratio of better than 40 dB and satisfies the potential target specifications of future mobile phones, extracted in this work.

This thesis studies the applicability of the electrocardiogram signal (ECG) as a biometric. There is strong evidence that heart's electrical activity embeds highly distinctive characteristics, suitable for applications such as the recognition of human subjects. Such systems traditionally provide two modes of functionality, identification and authentication; frameworks for subject recognition are herein proposed and analyzed in both scenarios. As in most pattern recognition problems, the probability of mis-classification error decreases as more learning information becomes available. Thus, a central consideration is the design and evaluation of algorithms which exploit the added information provided by the 12 lead standard ECG recording system. Feature and decision level fusion techniques described in thesis, offer enhanced security levels. The main novelty of the proposed approach, lies in the design of an identification system robust to cardiac arrhythmias. Criteria concerning the power distribution and information theoretic complexity of electrocardiogram windows are defined to signify abnormal ECG recordings, not suitable for recognition. Experimental results indicate high recognition rates and highlight identification based on ECG signals as very promising.

This work proposes a two-stage method that reconstructs the map of a scene from tagged photographs of that scene. In the first stage, several methods are proposed that transform tag data from the photographs into an intermediary distance matrix. These methods are compared against each other. In the second stage, an approach based on the physical mass-spring system is proposed that transforms the distance matrix into a map. This approach is compared against and outperforms MDS-MAP(P) when given human tagged input photographs. Experiments are carried out on two test datasets, one with 67 tags, and the other with 19. An evaluation method is described and the optimal overall reconstruction generates maps with accuracies of 47% and 66% respectively for the two test datasets, both scoring roughly 40% higher than a random reconstruction. The map reconstruction method is applied to three sample datasets and the resulting maps are qualitatively evaluated.

This thesis studies the applicability of two cardiac traits, the electrocardiogram (ECG) and the phonocardiogram (PCG), as biometrics. There is strong evidence that cardiac electrical activity (ECG) embeds highly distinctive characteristics, suitable for applications such as the recognition of human subjects. On the other hand, having the same origin with the ECG signal, it is believed that the PCG signal conveys distinctive information of an individual which can be deployed in biometric applications. Such recognition systems traditionally provide two modes of functionality, identification and authentication; frameworks for subject recognition are herein proposed and analyzed in both scenarios. Moreover, the expression of the cardiac signals is subject to alternation with heart rate and noise components. Thus, the central consideration of this thesis is the design and evaluation of robust recognition approaches that can compensate for these effects. A recognition system based on each, the ECG and the PCG, is developed and evaluated. Furthermore, a fusion of the two signals in a multimodal biometric system is investigated.