Semi-blind spatio-temporal equalization and multiuser detection for DS-CDMA systems

We propose a novel strategy for semi-blind spatio-temporal equalization and multiuser detection for short-burst asynchronous DS-CDMA systems. The technique is based on first performing semi-blind subspace-based channel identification and then semi-blind equalization via the constant modulus algorithm (CMA). In this way we are able to improve upon the traditional training-based least squares (LS) estimator by adding the constant modulus and cyclostationary properties of communication signals. Simulation results indicate a significant reduction in the required number of training symbols (for a short burst of data) compared with regularized LS estimators (both training-based and semi-blind). Application is in the reverse link of third-generation DS-CDMA systems.