Multiuser CDMA Signal Extraction

We consider the problem of extraction on wireless direct-sequence code-division multiple-access (DS-CDMA) systems. We develop an iterative least-squares-type algorithm that extracts the information symbols of the multiple concurrent DS-CDMA users and assumes no knowledge of either the signature waveforms or the channel state. The proposed scheme utilizes a novel initialization step that is based on auxiliary-vector (AV) subspace decomposition. Simulation studies evaluate the performance of the proposed extraction procedure in realistic data limited environments and show that, for observation records of sufficient length, information data extraction can be achieved with probability of error rather close to what may be attained with known signatures and channel state. *Corresponding author.