A survey of multiuser/multisubchannel detection schemes based on recurrent neural networks

Abstract

Multiuser/multisubchannel detection schemes based on a recurrent neural network (RNN) structure are a promising alternative to conventional detection schemes, both in terms of performance and complexity. This paper gives an overview of different approaches that make use of the results of RNN research for multiuser/multisubchannel detection. The detectors' results are compared with the results of conventional detection schemes in simulations of the uplink of a DS‐CDMA packet transmission system. Copyright © 2002 John Wiley & Sons, Ltd.

Soft iterative interference cancellation with successive over relaxation for digital transmission schemes based on multiple sets of orthogonal spreading codes

“…In this case VD with reasonable complexity and a performance as close as possible to the optimum must be applied. We use a VD which is based on soft-decision iterative interference cancellation [3], [7], [8]. This detector is very similar to the detector proposed by Vanhaverbeke et al [1].…”

“…using more codes than orthogonal spreading codes are available ( > 2 ) inescapably leads to interference between the subchannels due to the non-orthogonality of the codes. This interference must be combated by a suitable vector detection (VD) algorithm [3], [4].…”

“…Increasing the number of spreading codes will increase the bandwidth efficiency of the transmission. The interference between the subchannels resulting from nonorthogonal codes is removed by an iterative vector detection algorithm [3], [4]. We called this an m-OCDM ("m orthogonal code division multiplexing") transmission.…”

Vector equalization based on continuous-time recurrent neural networks

“…Iterative interference cancellation based on discrete-time RNNs without the need for training is a well investigated suboptimum scheme with moderate complexity and remarkable performance especially for channels with little to moderate interference [1], [10], [11], [12]. For properly defined RNN this can be explained by the equivalence between the energy function belonging to the RNN and the likelihood function in Eq.…”

Receiver Design for Uplink Multiuser Code Division Multiple Access Communication System Based on Neural Network

“…Neural networks have been employed extensively to solve a variety of difficult combinatorial optimization problems [7][8][9][10][11][12][13][19][20][21][22][23][24]. Next, we will transform the minimization of the likelihood function given in (26) into the minimization of neural network energy function E NN described by the expression (26) and (27), it is obvious that the minimization of neural network energy function and the minimization of the likelihood function are identical to each other.…”