Computationally Efficient Approaches for Blind Adaptive Beamforming in SIMO-OFDM Systems

In single-input multiple-output (SIMO) systems based on orthogonal frequency division multiplexing (OFDM), adaptive beamforming at the receiver side can be used to combat the effect of directional co-channel interference (CCI). Since pilot-aided beamforming suffers from consuming precious channel bandwidth, there has been much interest in blind beamforming approaches that can adapt their weights by restoring certain properties of the transmitted signals. Within this class of blind algorithms, the recursive least squares constant modulus algorithm (RLS-CMA) is of particular interest due to its good overall CCI cancelation performance and fast convergence. Nevertheless, the direct use of RSL-CMA within a SIMO-OFDM receiver induces considerable computational complexity, since a distinct copy of the RLS-CMA must be run on each individual sub-carriers. In this thesis, we present two approaches to reduce the computational complexity of SIMO-OFDM beamforming based on the RLS-CMA, namely: frequency interpolation and distributed processing. The former approach, which exploits the coherence bandwidth of the broadband wireless channels, divides the sub-carriers into several contiguous groups and applies the RLS-CMA to a selected sub-carrier in each group. The weight vectors at other frequencies are then obtained by interpolation. The distributed processing approach relies on the partitioning of the receiving array into sub-arrays and the use of a special approximation in the RLS-CMA. This allows a partial decoupling of the algorithm which can then be run on multiple processors with reduced overall complexity. This approach is well-suited to collaborative beamforming i~ multi-node distributed relaying. Through numerical simulation experiments of a SIMO-OFDM system, it is demonstrated that the proposed modifications to the RLS-CMA scheme can lead to substantial computational savings with minimal losses in adaptive cancelation performance.