Résumé: In mobile communication systems the space-time (S-T) channel matrix at the base station (BS) can be parameterized by a reduced set of unknowns with negligible distortion. In this correspondence, it is proposed to estimate the S-T channel under the constraint that the channel matrix is low-rank (the rank of the matrix accounts for the degrees of space and/or time diversity in the S-T channel). Low-rank properties of the S-T channel for a GSM/DCS system (rank is <2), prompted us to adapt the reduced-rank method for linear regression to estimate a rank-one channel. The proposed estimates coincide (or converge) with the maximum-likelihood (ML) approach when the training sequence is uncorrelated (or when the transmitted symbols are independent, i.e., in stochastic optimization). The beamforming weights and temporal channel are estimated jointly to reduce both cochannel interference (CCI) and intersymbol interference (ISI). The proposed receiver is easily adapted to the front-end architecture of the existing BS with minor modifications. The adaptive version of the algorithm is based on direct matrix inversion, guaranteeing convergence within a GSM burst and tracking time-varying channel and/or asynchronous interference.

Langue: Anglais