تجزیه و تحلیل عملکرد سنسورهای شبکه های عصبی در سیستم های DS / CDMA

In this paper, we consider neural networks as the detectors of signals of users in DS/CDMA systems. We apply multilayer perceptron neural network with back propagation learning algorithm in {AWGN} and multipath fading channels. Our analysis results in significant reduction in the receiver complexity over the previous studies. We compare the performance of neural network with the conventional and suboptimal detectors in {AWGN} channel and with the RAKE and single user lower bound receivers in fading channels. We also apply different criterion for training the network such as the decision based, fuzzy decision, discriminative learning, minimum classification, and entropy neural networks in {AWGN} channels and compare their performance. Further, we propose modified decision based network which improves the performance of the decision based network. A comparison between multilayer perceptron and Hopfield neural detectors is presented.