| The thesis addresses the problem of channel estimation in Impluse-Radio Ultra-Wideband (IR-UWB) communication system. The IR-UWB communications utilize low duty cycle pulses to transmit data over the wireless channel. The transmitted energy is distributed over a large number of multipath components (MPCs). At the receiver, these MPCs need to be estimated accurately to capture sufficient energy for successful communications. In our work, the IEEE 802.15.4a channel model is used where the channel is assumed to be Linear Time Invariant (LTI) and thus the problem of channel estimation becomes the estimation of the sparse channel taps and their delays. Since, the bandwidth of the signal is very large and the Nyquist rate sampling (∼ 16 GHz.) is impractical therefore we estimate the channel taps from the subsampled versions of the received signal profile. The transmitted pulse shape considered is the second derivative of the Gaussian pulse. We decompose the channel estimation problem into two parts: (i) estimation of the channel support, followed by, (ii) estimation of the support co-efficients (channel amplitudes). We exploting the signal sparsity and reduce the search space for the channel support by using three different methods: Genetic Algorithm, Correlation and Compressive Sensing. In the classical estimation approach we develop Low-Complexity Maximum Likelihood (LCML) estimator by leveraging the underlying structure of the problem. In the Bayesian framework, first we estimate the decomposed channel by incorporating the a priori multipath arrival time statistics for three different cases of amplitude statistics, namely (i) non-Gaussian, (ii) non-Gaussian with known second order statistics from the IEEE model, and (iii) Gaussian. Second, we jointly estimate the channel support and co-efficients by developing an Approximate Minimum Mean Square Error Estimator (AMMSE). We leverage the structure to reduce the computational complexity and propose a Low-Complexity MMSE (LCMMSE) channel estimator. The performance of the various methods in terms of the Normalized Root Mean Square Error (NRMSE) in estimation of MPC arrival times and energy capture were compared in the presence of AWGN. The novel low-complexity estimators, namely LCML, AMMSE and LCMMSE, presented in the thesis outperform other conventional UWB channel estimators. Furthermore, the computational complexity is much less as compared to that of Compressive Sensing, ML and MMSE estimators. |
