An adaptive neural network receiver for CDMA multi-user interference cancellation in multipath environments

The capacity of direct-sequence code-division multiple-access systems is severely limited by multiple access interference produced by the other co-channel users. To improve the capacity, joint-detection receivers and multi-access interference cancellation techniques have been proposed. Frequency-selective multi-path fading, however, has a significant effect on the performance of these types of receiver. This paper presents an adaptive neural-network-based receiver that has the capability of combating the `near-far´ problem and rejecting the multi-user interference in a multipath channel. A comparative performance analysis of the conventional matched-filter, a RAKE receiver and the adaptive neural network receiver is carried our using a Monte Carlo simulation. It is found that the new receiver performs extremely well as a suboptimal receiver in multipath and multiple access channels with a significant `near-far´ effect