State space based channel modelling

Abstract

xAbstract: An efficient channel estimation of single-input single output (SISO)/multiple-input multiple-output (MIMO) wireless channel is required for analysis, prediction, fault detection, channel equalization and optimization. This thesis presents a modified recursive MIMO Output-Error State Space Model Identification (MOESP) technique by using fast Givens transformations for LQ update and Lanczos algorithm with modified partial orthogonalization for singular value decomposition (SVD) update. It is based on one of the Subspace System Identification (SSI) technique known as MOESP method. Fast Givens transformation requires 4m multiplications which are less as compared to Givens rotations, as it requires 6m multiplications where m is size of the input vector. Lanczos algorithm with modified partial orthogonalization is applied thereby minimizing the number of vectors to be reorthogonalized for finding the basis for Krylov subspace in the SVD update. Complexity involved in modified partial orthogonalization is O(k:p)3. In existing recursive technique, more number of vectors needs to be orthogonalized by selective orthogonalization method according to the simulation results. Hence, proposed technique offers less complexity and is used for model identification of slow SISO/MIMO wireless channels.