Performance comparison of linear multiuser detectors and neural network detector for DS/CDMA systems in AWGN

Abstract

The most commonly used multiple access technique in wireless communication sphere is the direct sequence code division multiple access. The main drawback of this system is multiple access interference (MAI) caused by sharing a number of users the same channel. Multiuser Detection enhances the performance of DS-CDMA system by combating MAI. In this paper, the performance of the neural network detector is compared with the linear multiuser detectors includes decorrelating (Decor.) detector, and Minimum Mean Squared Error detector (MMSE). This neural network detects the user bits after the bank of matched filter in additive white Gaussian noise channel, with using spreading code of Gold sequence (GS) type. Where these detectors work in both synchronous and asynchronous transmission modes, in this paper its performance was investigated in synchronous AWGN channel. Simulation results show that the performance of linear multiuser detectors depends mainly on the number of active users. The neural network detector is superior to the linear multiuser detectors in the terms of bit error rate (BER) performance.