Convolutional recurrent neural network-based channel equalization

An experimental study

Yang Li, Minhua Chen, Yang Yang, Ming Tuo Zhou, Cheng-Xiang Wang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper, we revisit the idea of using deep neural network for channel equalization to account for nonlinear channel distortions as well as temporal variations of radio signals. Our insight is leveraging the the shift-invariant properties of the convolutional neural network (CNN) to learn matched filters analogous to the tap weights of conventional equalizer. Then we feed the learned filters into a subsequent recurrent neural network (RNN) with long-short-term-memory (LSTM) cells for temporal modeling of the channel. We train our proposed CNN-RNN (CRNN) equalizer based on real testbed collected data and enlarge the generalization ability of the learned network model as much as we can to adapt to different channel conditions. Experimental results show that the SER performance for our designated single-input single-output (SISO) system which utilises quadrature phase shift keying (QPSK) modulation scheme with the proposed CRNN-based channel equalizer outperforms that of other equalizers by average 2 to 5 dB at low signal-to-noise ratio (SNR).

Original languageEnglish
Title of host publication2017 23rd Asia-Pacific Conference on Communications (APCC)
PublisherIEEE
ISBN (Electronic)9781740523905
DOIs
Publication statusPublished - 1 Mar 2018

Fingerprint

Recurrent neural networks
Equalizers
Neural networks
Matched filters
Quadrature phase shift keying
Testbeds
Signal to noise ratio
Modulation

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Signal Processing

Cite this

Li, Y., Chen, M., Yang, Y., Zhou, M. T., & Wang, C-X. (2018). Convolutional recurrent neural network-based channel equalization: An experimental study. In 2017 23rd Asia-Pacific Conference on Communications (APCC) [8304090] IEEE. https://doi.org/10.23919/APCC.2017.8304090
Li, Yang ; Chen, Minhua ; Yang, Yang ; Zhou, Ming Tuo ; Wang, Cheng-Xiang. / Convolutional recurrent neural network-based channel equalization : An experimental study. 2017 23rd Asia-Pacific Conference on Communications (APCC). IEEE, 2018.
@inproceedings{19c67ebe15274294b8962875ed6ba890,
title = "Convolutional recurrent neural network-based channel equalization: An experimental study",
abstract = "In this paper, we revisit the idea of using deep neural network for channel equalization to account for nonlinear channel distortions as well as temporal variations of radio signals. Our insight is leveraging the the shift-invariant properties of the convolutional neural network (CNN) to learn matched filters analogous to the tap weights of conventional equalizer. Then we feed the learned filters into a subsequent recurrent neural network (RNN) with long-short-term-memory (LSTM) cells for temporal modeling of the channel. We train our proposed CNN-RNN (CRNN) equalizer based on real testbed collected data and enlarge the generalization ability of the learned network model as much as we can to adapt to different channel conditions. Experimental results show that the SER performance for our designated single-input single-output (SISO) system which utilises quadrature phase shift keying (QPSK) modulation scheme with the proposed CRNN-based channel equalizer outperforms that of other equalizers by average 2 to 5 dB at low signal-to-noise ratio (SNR).",
author = "Yang Li and Minhua Chen and Yang Yang and Zhou, {Ming Tuo} and Cheng-Xiang Wang",
year = "2018",
month = "3",
day = "1",
doi = "10.23919/APCC.2017.8304090",
language = "English",
booktitle = "2017 23rd Asia-Pacific Conference on Communications (APCC)",
publisher = "IEEE",
address = "United States",

}

Li, Y, Chen, M, Yang, Y, Zhou, MT & Wang, C-X 2018, Convolutional recurrent neural network-based channel equalization: An experimental study. in 2017 23rd Asia-Pacific Conference on Communications (APCC)., 8304090, IEEE. https://doi.org/10.23919/APCC.2017.8304090

Convolutional recurrent neural network-based channel equalization : An experimental study. / Li, Yang; Chen, Minhua; Yang, Yang; Zhou, Ming Tuo; Wang, Cheng-Xiang.

2017 23rd Asia-Pacific Conference on Communications (APCC). IEEE, 2018. 8304090.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Convolutional recurrent neural network-based channel equalization

T2 - An experimental study

AU - Li, Yang

AU - Chen, Minhua

AU - Yang, Yang

AU - Zhou, Ming Tuo

AU - Wang, Cheng-Xiang

PY - 2018/3/1

Y1 - 2018/3/1

N2 - In this paper, we revisit the idea of using deep neural network for channel equalization to account for nonlinear channel distortions as well as temporal variations of radio signals. Our insight is leveraging the the shift-invariant properties of the convolutional neural network (CNN) to learn matched filters analogous to the tap weights of conventional equalizer. Then we feed the learned filters into a subsequent recurrent neural network (RNN) with long-short-term-memory (LSTM) cells for temporal modeling of the channel. We train our proposed CNN-RNN (CRNN) equalizer based on real testbed collected data and enlarge the generalization ability of the learned network model as much as we can to adapt to different channel conditions. Experimental results show that the SER performance for our designated single-input single-output (SISO) system which utilises quadrature phase shift keying (QPSK) modulation scheme with the proposed CRNN-based channel equalizer outperforms that of other equalizers by average 2 to 5 dB at low signal-to-noise ratio (SNR).

AB - In this paper, we revisit the idea of using deep neural network for channel equalization to account for nonlinear channel distortions as well as temporal variations of radio signals. Our insight is leveraging the the shift-invariant properties of the convolutional neural network (CNN) to learn matched filters analogous to the tap weights of conventional equalizer. Then we feed the learned filters into a subsequent recurrent neural network (RNN) with long-short-term-memory (LSTM) cells for temporal modeling of the channel. We train our proposed CNN-RNN (CRNN) equalizer based on real testbed collected data and enlarge the generalization ability of the learned network model as much as we can to adapt to different channel conditions. Experimental results show that the SER performance for our designated single-input single-output (SISO) system which utilises quadrature phase shift keying (QPSK) modulation scheme with the proposed CRNN-based channel equalizer outperforms that of other equalizers by average 2 to 5 dB at low signal-to-noise ratio (SNR).

UR - http://www.scopus.com/inward/record.url?scp=85050598748&partnerID=8YFLogxK

U2 - 10.23919/APCC.2017.8304090

DO - 10.23919/APCC.2017.8304090

M3 - Conference contribution

BT - 2017 23rd Asia-Pacific Conference on Communications (APCC)

PB - IEEE

ER -

Li Y, Chen M, Yang Y, Zhou MT, Wang C-X. Convolutional recurrent neural network-based channel equalization: An experimental study. In 2017 23rd Asia-Pacific Conference on Communications (APCC). IEEE. 2018. 8304090 https://doi.org/10.23919/APCC.2017.8304090