Low complexity neural network structure for implementing the optimum maximum-likelihood multi-user receiver in a DS-CDMA communication system

The capacity of direct-sequence code division multiple access systems is interference limited, particularly by multiple-access interference produced by other co-channel users. The optimum multi-user receiver calculates the maximum-likelihood ratio of the detected data for all users simultaneously, but it has a complexity that grows exponentially with the number of users. In this paper, a neural network approach to multi-user detection is considered. It is shown that the performance of this receiver is the same as the maximum-likelihood multi-user receiver but it has a much lower computational complexity.