A gradient-based joint-detection algorithm for multi-carrier CDMA

In this paper, we investigate a gradient based joint detection algorithm for MC-CDMA. The algorithm is based on the improvement of a suboptimum initial estimation, which itself can be obtained from an already known low complexity detection algorithm. For the improvement strategy, the algorithm calculates the gradient of the Euclidean distance function between the rebuilt initially estimated transmission signal and the received signal and goes into the direction of the steepest descent in order to find an improved estimation, which is closer to the received signal. Simulation results show, that for 4-QAM modulation and a spreading length of 8, an error performance close to maximum likelihood detection can be achieved with a complexity which is about 26 times lower.