Description
TitleOn the performance of subspace SIMO blind channel identification methods
Date Created2017
Other Date2017-10 (degree)
Extent1 online resource (ix, 47 p. : ill.)
DescriptionChannel Identification is an important part of wireless communication systems. Radio-Frequency (RF) signals are subject to reflection, refraction, and diffraction, attenuation, and other effects, that result in a distorted signal at a receiver, particularly over what are known as frequency-selective channels. Traditionally, such distortion is estimated using a ``training sequence" which is a known reference signal used to estimate, and then correct for, the distortion. However, use of training sequences is not always possible, for example in military applications where the source signal is not known, or in broadcast environments where there is a high cost of transmitting a signal. One potential solution is to estimate the channel blindly, that is, without knowledge of the transmitted signal. Blind Channel Identification (BCI) and Equalization has been a extensive topic of research since at least 1975. One strategy in Blind Channel Identification is to use the structure of the received signals in a Single Input Multiple Output (SIMO) system to estimate the channel. Research has occurred on a number of methods that exploit this in the past several decades. The subspace methods form the channel estimate in terms of a one-dimensional subspace constructed using the estimated second-order statistics of the received signals. Additionally, the use of sparsity in signal estimation has been a topic of interest as well, and has recently been used in certain cases to improve the robustness of the subspace methods in a number of works. In this thesis, the Cross-Relations and Noise-Subspace methods, both of which are SIMO BCI methods, as well as their sparse variant, are examined for a deterministic channel. The expected Normalized Projection Misalignment (NPM) is analytically approximated for all considered methods. In addition, it is compared to simulation results for a random source signal and several measured RF channels from earlier literature. Finally, the sensitivity of the sparse variant of the subspace methods as a function of the regularization parameter is studied using simulation for a set of measured RF channels from earlier literature.
NoteM.S.
NoteIncludes bibliographical references
Noteby Kareem Y. Bonna
Genretheses, ETD graduate
Languageeng
CollectionSchool of Graduate Studies Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.
Statistical Profile