Deep-Learning-Based Signal Detection for Banded Linear Systems

Motivated by the recent advances in deep learning, we propose to design high-accuracy low-complexity signal detectors for banded linear systems based on deep neural networks (DNNs). We first design a fully connected DNN for signal detection. Then, to deal with the curse of dimensionality, we propose...

Full description

Saved in:
Bibliographic Details
Main Authors: Congmin Fan, Xiaojun Yuan, Zhang, Ying-Jun Angela
Format: Conference Proceeding
Language:English
Subjects:
Convolution
Deep learning
Detectors
Feature extraction
Neural networks
Neurons
Signal detection
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Motivated by the recent advances in deep learning, we propose to design high-accuracy low-complexity signal detectors for banded linear systems based on deep neural networks (DNNs). We first design a fully connected DNN for signal detection. Then, to deal with the curse of dimensionality, we propose a novel convolutional neural network (CNN) based on the banded structure of the channel matrix. From simulations, we observe that the proposed CNN outperforms the fully connected DNN in both accuracy and computational time. Moreover, CNN is more robust for the extension to channel matrices with a large size or a wide band. We also run extensive numerical experiments to show that both fully connected DNN and CNN perform much better than existing detectors with comparable complexity.
ISSN:2576-6813
DOI:10.1109/GLOCOM.2018.8648123