Blind FIR Channel Estimation in the Presence of Unknown Noise

Abstract

<p> In this thesis, we present three algorithms for blind estimation of the finite impulse response (FIR) channels in the presence of unknown noise. The algorithms are developed considering different available system resources: 1) If only one receiving antenna is available, based on the single-input-single-output (SISO) system model, with the output being up-sampled, we develop the maximum a posteriori (MAP) algorithm for Gaussian distributed noise. With large enough samples being collected, during which the channel keeps invariant, an efficient implementation of the MAP algorithm is also obtained; 2) If two receiving antennae can be affordable, based on the singleinput-multiple-output (SIMO) system model and up-sampling both the outputs, we develop a subspace based algorithm utilizing Canonical Correlation Decomposition (CCD) to obtain the subspaces, and a maximum likelihood (ML) based algorithm which starts from the Gaussian distributed projection error from the noise subspace onto the COD-estimated signal subspace. The developed channel estimators achieve superior performance measured by the normalized root mean square error (NRMSE), compared with some existing second-order-statistics (SOS) based methods while keeping the computation complexity comparable. When more than two receiving antennae are available, by treating them as one group and applying the MAP algorithm or separating them into two groups and applying the CCD based algorithms, the channels can still be blindly estimated with or without up-sampling the outputs. </p>