Symbol-aided channel estimation and multiuser detection for CDMA systems using a decorrelating detector

A decorrelating filter detector is considered for multiuser detection. This filter operates on the outputs of matched filters which are matched to the original code waveforms of system. A maximum likelihood estimation to estimate the parameters of the decorrelating filter is proposed and derived. The estimation method is based on inserting known trailing sequences into the information data by all users simultaneously. To achieve the minimum mean square error (MMSE) in the estimation, a criterion for selecting the training sequences is suggested. Orthogonal training sequences are good candidates to approach the MMSE. The estimation method requires a matrix inversion at the end of each training period. An iterative matrix inversion method is suggested to distribute the computational load of the matrix inversion over the training period. The simulation results show some degradations in system performance compared to the case where a perfect knowledge of the channel is assumed. This degradation is clearly due to errors in the channel estimation.