Computationally efficient multiuser detection for aperiodically spreading CDMA systems

Multiuser CDMA systems with aperiodic spreading codes have received considerable attention. One major difficulty in multiuser detection is the user´s time-varying signature. In a multipath communication environment, this signature is not a priori known, making direct design of the detector more intractable. Multiuser detection is performed by a two-step approach: estimate the unknown channel coefficients and construct the detector based on the MMSE criterion. We model the interference as colored Gaussian noise with unknown correlations and estimate the channel parameters based on the correlation matching idea. The spreading codes are assumed random with known up to the fourth order statistics, resulting in a computationally efficient method. In the case of unknown statistics however, the method is still applicable by estimating those statistics from given codes. To guarantee identifiability, both the channel´s output and the output of the matched filter for the desired user are considered in one cost function. Comparisons with a previously proposed method and other existing methods are made by simulations