Analysis of a bayesian multiuser. detector for non-data-aided cdma communications

This paper presents a Bayesian multiuser detection (B-MUD) algorithm, based on the extension of previous work of the authors on single-user equalization via RBF/neural networks-based equalizers. Since all Bayesian equalizing algorithms have an inherent complexity that increases exponentially with time, the proposed MUD introduces a novel complexity-controlling scheme, in order to keep the computational load in an admissible level. One of the main advantages of this Bayesian approach is its capability to estimate the bit error rate in a non-data-aided mode. A detailed analysis of the influence on the performance of the most relevant parameters, such as the orthogonality of the signatures, the processing gain or the near-far effect, is carried out in the simulations section