A State-Space Approach to Semi-Blind Adaptive Multiuser Detection in Time-Varying Environment

Autoregressive (AR) model derived from Jakes´ Doppler Spectrum is used for signature sequence estimation in direct sequence code division multiple access (DS-CDMA) system. A decision directed algorithm is proposed to work on iterative basis during the course of data transmission in a rapidly time-varying environment. The algorithm is based on state-space approach which uses Kalman filter for signature sequence estimation in the presence of multipath and time varying channel. The proposed algorithm successfully works in tracking rapid changes of the channel whereas gradient-based algorithm like LMS almost fails in doing so. A linear minimum mean squared error (MMSE) detector is used to detect the symbols, which are then used in updating the received signature waveforms, distorted by the time-varying channel. Simulation results are presented which show that the performance of a linear adaptive receiver can be improved significantly with signature tracking on high Doppler spreads in DS-CDMA system