Semiblind maximum-likelihood demodulation for CDMA systems

This correspondence addresses the problem of channel estimation and symbol detection in wireless direct-sequence code-division multiple-access (DS-CDMA) communication systems. We introduce a novel multiuser demodulation scheme that proceeds in two steps. First, the multiaccess channel parameters are estimated according to a suitable modification of the maximum-likelihood (ML) criterion using the expectation maximization (EM) algorithm. Subsequently, this estimate and other useful side information are employed to perform ML detection of the transmitted symbols with the Viterbi algorithm. Our main contribution is the development of a novel stochastic ML method for channel estimation that takes advantage of all the available statistical information referred to the transmitted signals and channel noise. Additionally, it can incorporate the knowledge of a fraction of the transmitted symbols; hence, the term semiblind. Computer simulation results are presented that show how close-to-optimum performance is achieved in time-dispersive fading channels using remarkably short training sequences.