Multistage semi-blind spatio-temporal processing for short burst multiuser SDMA systems

In short burst multiuser Spatial Division Multiple Access (SDMA) systems, the accessible volume of data for adjusting the coefficients of spatio-temporal filters may not be large enough to allow global convergence of existing semi-blind algorithms. A multistage algorithm which significantly increases the probability of global convergence without enlarging the training sequence is presented. It is based on the observation that in multiuser environments, it may be efficient to achieve global convergence for "at least one user" rather than global convergence for "all users". Then, information from one stage can be used in the next stage to enhance the estimation of the remaining users.