Theses and Dissertations from UMD
Permanent URI for this communityhttp://hdl.handle.net/1903/2
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a give thesis/dissertation in DRUM
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
41 results
Search Results
Item Spectral LADAR: Active Range-Resolved Imaging Spectroscopy(2011) Powers, Michael; Davis, Christopher C; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Imaging spectroscopy using ambient or thermally generated optical sources is a well developed technique for capturing two dimensional images with high per-pixel spectral resolution. The per-pixel spectral data is often a sufficient sampling of a material's backscatter spectrum to infer chemical properties of the constituent material to aid in substance identification. Separately, conventional LADAR sensors use quasi-monochromatic laser radiation to create three dimensional images of objects at high angular resolution, compared to RADAR. Advances in dispersion engineered photonic crystal fibers in recent years have made high spectral radiance optical supercontinuum sources practical, enabling this study of Spectral LADAR, a continuous polychromatic spectrum augmentation of conventional LADAR. This imaging concept, which combines multi-spectral and 3D sensing at a physical level, is demonstrated with 25 independent and parallel LADAR channels and generates point cloud images with three spatial dimensions and one spectral dimension. The independence of spectral bands is a key characteristic of Spectral LADAR. Each spectral band maintains a separate time waveform record, from which target parameters are estimated. Accordingly, the spectrum computed for each backscatter reflection is independently and unambiguously range unmixed from multiple target reflections that may arise from transmission of a single panchromatic pulse. This dissertation presents the theoretical background of Spectral LADAR, a shortwave infrared laboratory demonstrator system constructed as a proof-of-concept prototype, and the experimental results obtained by the prototype when imaging scenes at stand off ranges of 45 meters. The resultant point cloud voxels are spectrally classified into a number of material categories which enhances object and feature recognition. Experimental results demonstrate the physical level combination of active backscatter spectroscopy and range resolved sensing to produce images with a level of complexity, detail, and accuracy that is not obtainable with data-level registration and fusion of conventional imaging spectroscopy and LADAR. The capabilities of Spectral LADAR are expected to be useful in a range of applications, such as biomedical imaging and agriculture, but particularly when applied as a sensor in unmanned ground vehicle navigation. Applications to autonomous mobile robotics are the principal motivators of this study, and are specifically addressed.Item A STUDY ON DISTRIBUTED RECEDING HORIZON CONTROL(2011) MA, WANN-JIUN; Martins, Nuno; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)We consider a distributed control problem comprising of multiple sub-systems with one-controller at each sub-system. We apply a recent result about suboptimal receding horizon control that analytically relates a receding horizon control suboptimal solution and system performance loss to quantify the necessary number of iterations for the dual and primal decomposition algorithm to achieve a solution that guarantees stability. We also use this result to explore the idea of "incremental robustness", meaning that the overall system is robustly stable and its performance varies gracefully with the inclusion of sub-systems and sub-controllers. We demonstrate these ideas in a consensus seeking and a formation control problem and provide simulation results. To our best knowledge, this is the first time the result is applied to a distributed receding horizon control framework based on dual and primal decomposition.Item Multimedia Protection using Content and Embedded Fingerprints(2011) Varna, Avinash Laxmisha; Wu, Min; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Improved digital connectivity has made the Internet an important medium for multimedia distribution and consumption in recent years. At the same time, this increased proliferation of multimedia has raised significant challenges in secure multimedia distribution and intellectual property protection. This dissertation examines two complementary aspects of the multimedia protection problem that utilize content fingerprints and embedded collusion-resistant fingerprints. The first aspect considered is the automated identification of multimedia using content fingerprints, which is emerging as an important tool for detecting copyright violations on user generated content websites. A content fingerprint is a compact identifier that captures robust and distinctive properties of multimedia content, which can be used for uniquely identifying the multimedia object. In this dissertation, we describe a modular framework for theoretical modeling and analysis of content fingerprinting techniques. Based on this framework, we analyze the impact of distortions in the features on the corresponding fingerprints and also consider the problem of designing a suitable quantizer for encoding the features in order to improve the identification accuracy. The interaction between the fingerprint designer and a malicious adversary seeking to evade detection is studied under a game-theoretic framework and optimal strategies for both parties are derived. We then focus on analyzing and understanding the matching process at the fingerprint level. Models for fingerprints with different types of correlations are developed and the identification accuracy under each model is examined. Through this analysis we obtain useful guidelines for designing practical systems and also uncover connections to other areas of research. A complementary problem considered in this dissertation concerns tracing the users responsible for unauthorized redistribution of multimedia. Collusion-resistant fingerprints, which are signals that uniquely identify the recipient, are proactively embedded in the multimedia before redistribution and can be used for identifying the malicious users. We study the problem of designing collusion resistant fingerprints for embedding in compressed multimedia. Our study indicates that directly adapting traditional fingerprinting techniques to this new setting of compressed multimedia results in low collusion resistance. To withstand attacks, we propose an anti-collusion dithering technique for embedding fingerprints that significantly improves the collusion resistance compared to traditional fingerprints.Item Combinatorial Methods in Coding Theory(2011) Mazumdar, Arya; Barg, Alexander; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This thesis is devoted to a range of questions in applied mathematics and signal processing motivated by applications in error correction, compressed sensing, and writing on non-volatile memories. The underlying thread of our results is the use of diverse combinatorial methods originating in coding theory and computer science. The thesis addresses three groups of problems. The first of them is aimed at the construction and analysis of codes for error correction. Here we examine properties of codes that are constructed using random and structured graphs and hypergraphs, with the main purpose of devising new decoding algorithms as well as estimating the distribution of Hamming weights in the resulting codes. Some of the results obtained give the best known estimates of the number of correctable errors for codes whose decoding relies on local operations on the graph. In the second part we address the question of constructing sampling operators for the compressed sensing problem. This topic has been the subject of a large body of works in the literature. We propose general constructions of sampling matrices based on ideas from coding theory that act as near-isometric maps on almost all sparse signal. This matrices can be used for dimensionality reduction and compressed sensing. In the third part we study the problem of reliable storage of information in non-volatile memories such as flash drives. This problem gives rise to a writing scheme that relies on relative magnitudes of neighboring cells, known as rank modulation. We establish the exact asymptotic behavior of the size of codes for rank modulation and suggest a number of new general constructions of such codes based on properties of finite fields as well as combinatorial considerations.Item Single-Event Gate Rupture in Power MOSFETs: A New Radiation Hardness Assurance Approach(2011) Lauenstein, Jean-Marie; Goldsman, Neil; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Almost every space mission uses vertical power metal-semiconductor-oxide field-effect transistors (MOSFETs) in its power-supply circuitry. These devices can fail catastrophically due to single-event gate rupture (SEGR) when exposed to energetic heavy ions. To reduce SEGR failure risk, the off-state operating voltages of the devices are derated based upon radiation tests at heavy-ion accelerator facilities. Testing is very expensive. Even so, data from these tests provide only a limited guide to on-orbit performance. In this work, a device simulation-based method is developed to measure the response to strikes from heavy ions unavailable at accelerator facilities but posing potential risk on orbit. This work is the first to show that the present derating factor, which was established from non-radiation reliability concerns, is appropriate to reduce on-orbit SEGR failure risk when applied to data acquired from ions with appropriate penetration range. A second important outcome of this study is the demonstration of the capability and usefulness of this simulation technique for augmenting SEGR data from accelerator beam facilities. The mechanisms of SEGR are two-fold: the gate oxide is weakened by the passage of the ion through it, and the charge ionized along the ion track in the silicon transiently increases the oxide electric field. Most hardness assurance methodologies consider the latter mechanism only. This work demonstrates through experiment and simulation that the gate oxide response should not be neglected. In addition, the premise that the temporary weakening of the oxide due to the ion interaction with it, as opposed to due to the transient oxide field generated from within the silicon, is validated. Based upon these findings, a new approach to radiation hardness assurance for SEGR in power MOSFETs is defined to reduce SEGR risk in space flight projects. Finally, the potential impact of accumulated dose over the course of a space mission on SEGR susceptibility is explored. SEGR evaluation of gamma-irradiated power MOSFETs suggests a non-significant SEGR susceptibility enhancement due to accumulated dose from gamma rays. During SEGR testing, an unexpected enhanced dose effect from heavy-ion irradiation was detected. We demonstrate that this effect could be due to direct ionization by two or more ions at the same channel location. The probability on-orbit for such an occurrence is near-zero given the low heavy-ion fluence over a typical mission lifetime, and did not affect SEGR susceptibility. The results of this work can be used to bound the risk of SEGR in power MOSFETs considered for insertion into spacecraft and instruments.Item WIRELESS NETWORK COCAST: COOPERATIVE COMMUNICATIONS WITH SPACE-TIME NETWORK CODING(2011) Lai, Hung-Quoc Duc; Liu, K. J. Ray; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Traditional cooperative communications can greatly improve communication performance. However, transmissions from multiple relay nodes are challenging in practice. Single transmissions using time-division multiple access cause large transmission delay, but simultaneous transmissions from two or more nodes using frequency-division multiple access (FDMA), code-division multiple access (CDMA), or distributed space-time codes are associated with the issues of imperfect frequency and timing synchronization due to the asynchronous nature of cooperation. In this dissertation, we propose a novel concept of wireless network cocast (WNC) and develop its associated space-time network codes (STNCs) to overcome the foretold issues. In WNC networks, each node is allocated a time slot for its transmission and thus the issues of imperfect synchronization are eliminated. To reduce the large transmission delay, each relay node forms a unique signal, a combination of the overheard information, and transmits it to the intended destination. The combining functions at relay nodes form a STNC that ensures full spatial diversity for the transmitted information as in traditional cooperative communications. Various traditional combining techniques are utilized to design the STNCs, including FDMA-like and CDMA-like techniques and transform-based techniques with the use of Hadamard and Vandermonde matrices. However, a major distinction is that the combination of information from different sources happens within a relay node instead of through the air as in traditional cooperative communications. We consider a general case of multiuser relay wireless networks, where user nodes transmit and receive their information to and from a common base node with the assistance from relay nodes. We then apply the STNCs to multiuser cooperative networks, in which the user nodes are also relay nodes helping each other in their transmission. Since the cooperative nodes are distributed around the network, the node locations can be an important aspect of designing a STNC. Therefore, we propose a location-aware WNC scheme to reduce the aggregate transmit power and achieve even power distribution among the user nodes in the network. WNC networks and its associated STNCs provide spatial diversity to dramatically reduce the required transmit power. However, due to the additional processing power in receiving and retransmitting each other's information, not all nodes and WNC networks result in energy efficiency. Therefore, we first examine the power consumption in WNC networks. We then offer a TDMA-based merge process based on coalitional formation games to orderly and efficiently form cooperative groups in WNC networks. The proposed merge process substantially reduces the network power consumption and improves the network lifetime.Item NOVEL TUNNELING DIODES FOR A HIGH PERFORMANCE INFRARED RECTENNA(2011) Choi, Kwangsik; Peckerar, Martin; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Since the 1960s, metal-insulator-metal (MIM) tunneling diodes have been used for detecting and mixing electromagnetic waves up to infrared frequencies. To improve the wave coupling efficiency, an antenna is usually integrated with a MIM diode, and this integrated structure is known as a "rectenna" which can rectify incident waves. Although antenna coupled MIM diodes can detect and rectify infrared waves, the energy conversion efficiency of these structures is usually very low largely because of the response time of the tunnel junction. This thesis summarizes an attempt to improve the power conversion efficiency of the rectenna. As a result, novel tunneling diodes have been developed using a geometric field enhancement (GFE) technique, which takes advantage of the "lightning rod" effect. The GFE technique is implemented by using a pointed electrode, creating an asymmetric electric field in the region of the tunnel barrier. Thus, the tunneling current in this system can be asymmetric with respect to the sign of the applied bias. Furthermore, the geometric structure of these novel optical rectennas provides the appropriate conditions for the excitation of surface plasmon resonances. These resonances further enhance the junction's electric field, allowing for larger current flow, effectively lowering the diodes tunneling resistance. Three different types of tunneling diodes are developed and explored in this research: a perfect planar type tunneling diode, an asymmetric tunneling diode (ATD), and a focused asymmetric metal-oxide-metal (FAMIM) tunneling diode. The fabrication processes for each new tunneling diode have been successfully developed. The degree of performance improvement achieved by each process is summarized. This thesis documents the highest MIM detection sensitivities of 31 V−1 and 22 V−1 reported in open literature for planar type tunneling diodes and ATDs, respectively. Improvements in tunneling current nonlinearity (curvature of the current-voltage plot) of 350 % and 33 % are achieved for ATDs and FAMIM diodes, respectively.Item SEGREGATION OF SPEECH SIGNALS IN NOISY ENVIRONMENTS(2011) Vishnubhotla, Srikanth; Espy-Wilson, Carol Y; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Automatic segregation of overlapping speech signals from single-channel recordings is a challenging problem in speech processing. Similarly, the problem of extracting speech signals from noisy speech is a problem that has attracted a variety of research for several years but is still unsolved. Speech extraction from noisy speech mixtures where the background interference could be either speech or noise is especially difficult when the task is to preserve perceptually salient properties of the recovered acoustic signals for use in human communication. In this work, we propose a speech segregation algorithm that can simultaneously deal with both background noise as well as interfering speech. We propose a feature-based, bottom-up algorithm which makes no assumptions about the nature of the interference or does not rely on any prior trained source models for speech extraction. As such, the algorithm should be applicable for a wide variety of problems, and also be useful for human communication since an aim of the system is to recover the target speech signals in the acoustic domain. The proposed algorithm can be compartmentalized into (1) a multi-pitch detection stage which extracts the pitch of the participating speakers, (2) a segregation stage which teases apart the harmonics of the participating sources, (3) a reliability and add-back stage which scales the estimates based on their reliability and adds back appropriate amounts of aperiodic energy for the unvoiced regions of speech and (4) a speaker assignment stage which assigns the extracted speech signals to their appropriate respective sources. The pitch of two overlapping speakers is extracted using a novel feature, the 2-D Average Magnitude Difference Function, which is also capable of giving a single pitch estimate when the input contains only one speaker. The segregation algorithm is based on a least squares framework relying on the estimated pitch values to give estimates of each speaker's contributions to the mixture. The reliability block is based on a non-linear function of the energy of the estimates, this non-linear function having been learnt from a variety of speech and noise data but being very generic in nature and applicability to different databases. With both single- and multiple- pitch extraction and segregation capabilities, the proposed algorithm is amenable to both speech-in-speech and speech-in-noise conditions. The algorithm is evaluated on several objective and subjective tests using both speech and noise interference from different databases. The proposed speech segregation system demonstrates performance comparable to or better than the state-of-the-art on most of the objective tasks. Subjective tests on the speech signals reconstructed by the algorithm, on normal hearing as well as users of hearing aids, indicate a significant improvement in the perceptual quality of the speech signal after being processed by our proposed algorithm, and suggest that the proposed segregation algorithm can be used as a pre-processing block within the signal processing of communication devices. The utility of the algorithm for both perceptual and automatic tasks, based on a single-channel solution, makes it a unique speech extraction tool and a first of its kind in contemporary technology.Item CMOS IMAGE SENSORS FOR LAB-ON-A-CHIP MICROSYSTEM DESIGN(2011) Sander, David; Abshire, Pamela; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)The work described herein serves as a foundation for the development of CMOS imaging in lab-on-a-chip microsystems. Lab-on-a-chip (LOC) systems attempt to emulate the functionality of a cell biology lab by incorporating multiple sensing modalidites into a single microscale system. LOC are applicable to drug development, implantable sensors, cell-based bio-chemical detectors and radiation detectors. The common theme across these systems is achieving performance under severe resource constraints including noise, bandwidth, power and size. The contributions of this work are in the areas of two core lab-on-a-chip imaging functions: object detection and optical measurements.Item MODELING AND TESTING OF ETHERNET TRANSFORMERS(2011) Bowen, David; Mayergoyz, Isaak D; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Twisted-pair Ethernet is now the standard home and office last-mile network technology. For decades, the IEEE standard that defines Ethernet has required electrical isolation between the twisted pair cable and the Ethernet device. So, for decades, every Ethernet interface has used magnetic core Ethernet transformers to isolate Ethernet devices and keep users safe in the event of a potentially dangerous fault on the network media. The current state-of-the-art Ethernet transformers are miniature (<5mm diameter) ferrite-core toroids wrapped with approximately 10 to 30 turns of wire. As small as current Ethernet transformers are, they still limit further Ethernet device miniaturization and require a separate bulky package or jack housing. New coupler designs must be explored which are capable of exceptional miniaturization or on-chip fabrication. This dissertation thoroughly explores the performance of the current commercial Ethernet transformers to both increase understanding of the device's behavior and outline performance parameters for replacement devices. Lumped element and distributed circuit models are derived; testing schemes are developed and used to extract model parameters from commercial Ethernet devices. Transfer relation measurements of the commercial Ethernet transformers are compared against the model's behavior and it is found that the tuned, distributed models produce the best transfer relation match to the measured data. Process descriptions and testing results on fabricated thin-film dielectric-core toroid transformers are presented. The best results were found for a 32-turn transformer loaded with 100Ω, the impedance of twisted pair cable. This transformer gave a flat response from about 10MHz to 40MHz with a height of approximately 0.45. For the fabricated transformer structures, theoretical methods to determine resistance, capacitance and inductance are presented. A special analytical and numerical analysis of the fabricated transformer inductance is presented. Planar cuts of magnetic slope fields around the dielectric-core toroid are shown that describe the effect of core height and winding density on flux uniformity without a magnetic core.