An application of the maximum likelihood principle to semiblind space-time linear detection in multiple-access wireless communications

This paper introduces a novel semiblind approach to space-time linear detection in multiple-access systems. A new criterion for the selection of the linear receiver coefficients, based on the maximum likelihood (ML) principle, is derived and a practical implementation by means of a fast expectation-maximization (EM) algorithm is suggested. The semiblind criterion is obtained from a purely statistical point of view where the aim of training data is not to enhance performance but to eliminate misconvergence problems