Show simple item record

AuthorSparrer, Susannedc.contributor.author
Date of accession2020-08-26T10:44:40Zdc.date.accessioned
Available in OPARU since2020-08-26T10:44:40Zdc.date.available
Year of creation2019dc.date.created
Date of first publication2020-08-26dc.date.issued
AbstractIn the multiple-input/multiple-output scenario which is an important field of research in communications engineering, the data transmission from a couple of transmit antennas to a number of receive antennas is investigated, where usually the number of receivers is not smaller than the number of transmitters. In some applications, however, only a small number of the transmitters is active, allowing for a reliable transmission even if the number of transmitters exceeds the number of receivers. In order to determine which transmitters where active in a certain time slot and which data has been transmitted by them, a sparse vector has to be reconstructed from an underdetermined system of linear equations. This problem is denoted as compressed sensing, where the signal to be estimated is usually real-valued (or in same cases complex-valued, which is also the standard multiple-input/multiple-output setup). In many applications such as digital data transmission, however, the signal to be recovered is discrete-valued. Moreover, in every real world scenario the measurements are corrupted by additive noise which complicates the reconstruction of the original signal. In this thesis, the estimation of discrete-valued sparse signals from underdetermined linear systems in the presence of additive noise, the so-called discrete compressed sensing, is investigated. To this end, as discrete compressed sensing combines two main areas of communications engineering, namely, the joint estimation of multiple transmitters and the processing of discrete-valued data, knowledge and techniques known from communications engineering are applied, complementing the mainly mathematical perspective of the research on standard compressed sensing. In the first part (Chapter 3), the estimation principles of interest are discussed from a theoretical point of view. To this end, the different constraints which are to be taken into account are considered, and strategies for the estimation of the signal with respect to the channel model at hand on the one side, and with respect to the sparsity constraint and the discrete nature of the signal on the other side, are discussed. Moreover, the unbiasing of linear as well as nonlinear estimators is analyzed. In the second part (Chapters 4 and 5) of this thesis, based on standard algorithms for compressed sensing we develop powerful well-performing algorithms, which take into account the channel model including the additive noise, as well as the knowledge on the discrete setup and the sparsity. Therefore, the estimation principles utilized in the algorithms given in the literature are reviewed and replaced by superior approaches which were presented in the first part. Adaptations are discussed for successive and for iterative algorithms. The superiority of the algorithms developed in this thesis is demonstrated via numerical simulations. Furthermore, the resulting algorithms are analyzed from a communications engineering point of view, and compared to turbo-based estimation and message passing, which are well-known in channel coding. The in-depth study of the similarities and differences and the underlying approaches and approximations leads to an extensive understanding of the algorithms. The advantages and disadvantages of (approximate) message passing and turbo-based estimation are elucidated, and the equivalence of unbiasing and the calculation of extrinsic values is shown. Note that, although the focus of this thesis is on discrete compressed sensing, the proposed algorithms can be applied for every desired setup in which the a-priori probability density function of the sparse signal is known.dc.description.abstract
Languageendc.language.iso
PublisherUniversität Ulmdc.publisher
LicenseStandarddc.rights
Link to license texthttps://oparu.uni-ulm.de/xmlui/license_v3dc.rights.uri
KeywordUnbiasingdc.subject
KeywordDiscrete compressed sensingdc.subject
KeywordMMSE estimationdc.subject
KeywordApproximate message passingdc.subject
KeywordOrthogonal matching pursuitdc.subject
KeywordIterative hard thresholdingdc.subject
KeywordSoft feedbackdc.subject
KeywordTurbo principledc.subject
KeywordRobust compressed sensingdc.subject
KeywordCompressive samplingdc.subject
Dewey Decimal GroupDDC 300 / Social sciencesdc.subject.ddc
Dewey Decimal GroupDDC 600 / Technology (Applied sciences)dc.subject.ddc
LCSHCompressed sensing (Telecommunication)dc.subject.lcsh
LCSHMIMO systemsdc.subject.lcsh
LCSHDigital communicationsdc.subject.lcsh
TitleAlgorithms for (discrete) compressed sensing : a communications engineering perspectivedc.title
Resource typeDissertationdc.type
Date of acceptance2019-09-17dcterms.dateAccepted
RefereeFischer, Robertdc.contributor.referee
RefereeGörtz, Norbertdc.contributor.referee
DOIhttp://dx.doi.org/10.18725/OPARU-32683dc.identifier.doi
PPN1728476909dc.identifier.ppn
URNhttp://nbn-resolving.de/urn:nbn:de:bsz:289-oparu-32745-7dc.identifier.urn
GNDMIMOdc.subject.gnd
GNDNachrichtentechnikdc.subject.gnd
GNDKomprimierte Abtastungdc.subject.gnd
GNDSchwellwertverfahrendc.subject.gnd
FacultyFakultät für Ingenieurwissenschaften, Informatik und Psychologieuulm.affiliationGeneral
InstitutionInstitut für Nachrichtentechnikuulm.affiliationSpecific
Grantor of degreeFakultät für Ingenieurwissenschaften, Informatik und Psychologieuulm.thesisGrantor
DCMI TypeTextuulm.typeDCMI
CategoryPublikationenuulm.category
FundingDiskret-wertige dünn besetzte Signale: Theorie, Algorithmen und Anwendungen / DFG [257184199, FI 982/8-1]uulm.funding
Bibliographyuulmuulm.bibliographie


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record