Optimum Multiuser Detection for CDMA Systems Using the Mean Field Annealing Neural Network

Abstract

In this chapter, we investigate the application of Mean Field Annealing (MFA) neural networks to the optimum multiuser detection in a direct-sequence code-division multiple-access (DS/CDMA) system over the additive white Gaussian noise (AWGN) channel. Although the optimum receiver for multiuser detection is superior to the conventional matched filter receiver when the relative powers of the interfering signals are large, the optimum receiver obtained by the maximization of a log-likelihood function has a complexity that's is exponential in the number of users. This prohibitive complexity has spawned the area of research on employing neural network techniques to develop an optimum detector with moderate complexity. In this chapter, it is shown that the energy function of the neural network can be derived from and is then expressed in terms of the likelihood function of the optimum multiuser detection for both the synchronous and asynchronous CDMA systems. An MFA network, which combines characteristics of the simulated annealing algorithm and the Hopfield neural network is proposedto seek out the global optimum solution of this energy function. Additionally, MFA exhibits the rapid convergence of the neural network while preserving the solution quality afforded by the stochastic simulated annealing algorithm. This would lead to a cost-effective and efficient minimization mechanism for CDMA multiuser detection. Computer simulation carry out performance comparisons among optimum detection, matched filter detection and MFA detection.