ML Detection of MIMO-OFDM Signals in Selected Spatial-Temporal Subspace for Prewhitening with Recursive Eigenvalue Decomposition in Mobile Interference Environments

This paper proposes a maximum likelihood (ML) detector with spatial-temporal filters prewhitening cochannel interferences for MIMO-OFDM mobile communications. The proposed detector consists of several branch metric generators and an ML detector using metrics which the generators produce. In the branch metric generator, the spatial-temporal filter for received signals suppresses interfering signal components while prewhitening them. Subtracting replica signals from the filter´s outputs and squaring the result yield the metrics. Coefficients of the filters and channel impulse responses for the replicas are estimated by using the eigenvalue decomposition (EVD) of an autocorrelation matrix of both the received and transmitted signals. To select appropriate eigenvectors for the branch metric generators, the proposed detector adopts a channel capacity criterion, and avoids prohibitive complexity by using a fast search algorithm. Another problem is that it requires a large amount of computational complexity for channel tracking because it performs parameter estimation using the nonrecursive EVD. To solve such a problem, the recursive EVD is applied, which can track the time-varying fading channel while simultaneously reducing the complexity. Computer simulations demonstrate that the proposed scheme can maintain an excellent BER performance in correlated channels and track the fast fading channel with reduced complexity.