Summary
The 2018 International Symposium on Information Theory and Its Applications　(ISITA2018)
2018
Session Number:We-AM-Poster
Session:
Number:We-AM-Poster.15
Neural Network Detection of LDPC-Coded Random Access CDMA Systems
Yuto Ichiki, Guanghui Song, Kui Cai, Shan Lu, Jun Cheng,
pp.505-505
Publication Date:2018/10/18
Online ISSN:2188-5079
DOI:10.34385/proc.55.We-AM-Poster.15
PDF download
Summary:
Neural network-based CDMA detection is previously shown with near optimal performance and complexity lower than conventional detection schemes [1][2]. However, these works considered uncoded CDMA with a determinate communication user number. The robustness of the neural network multiuser detection (NNMD) in coded CDMA with random user number, and the problem of how to implement soft-decision channel decoding in this system is yet to be investigated. In this work, we consider LDPC-coded CDMA system with random number of active users and propose an NNMD to detect multiple users’ activities and data of active users. The NNMD provides a soft estimation for each of the LDPC coded bit, based on which, a soft LDPC decoding is implemented. Both the user activity identification and LDPC decoding are based on a statistical probability density function (PDF) of the NNMD output. Numerical results validate the system robustness to random access and channel perturbations.