Low complexity linear detection for asynchronous CDMA signals over frequency selective channels

In this paper, a suboptimal low complexity structure for joint linear detection of CDMA signals is proposed. The optimal joint linear zero-forcing or minimum mean squared error detectors use a vector model of the transmission where the channel is represented by a matrix. Due to the finite memory of the channel, most of the blocks of the channel autocorrelation matrix are equal to zero. It is then possible to put this structure into profit to derive low complexity implementations for these detectors. In this paper, the optimal linear system resolution is approximated by overlapped smaller subsystems. The performance and the complexity of the resulting detector are analyzed for both downlink and uplink communications over frequency selective channels for periodic spreading codes with a period equal to the symbol period as well as for random-like codes. The trade-off between complexity and performance is flexible, which offers a great advantage to the proposed solution.