Deep-learning neural network for MIMO detection in a mode-division multiplexed optical transmission system

被引：0

作者：

Poudel, Bishal ^{[1
]}

Oshima, Joji ^{[1
]}

Kobayashi, Hirokazu ^{[1
]}

Iwashita, Katsushi ^{[1
]}

机构：

[1] Kochi Univ Technol, Dept Photon Syst Engn, 185 Miyanokuchi, Kochi 7828502, Japan

来源：

NEXT-GENERATION OPTICAL COMMUNICATION: COMPONENTS, SUB-SYSTEMS, AND SYSTEMS VIII | 2019年 / 10947卷

关键词：

Mode division multiplexing; multiple input multiple output; multimode fiber; Deep learning neural network; TensorFlow; Keras library; MIMO detectors;

D O I：

10.1117/12.2505445

中图分类号：

O43 [光学];

学科分类号：

070207 ; 0803 ;

摘要：

In this paper, a Mode division multiplexing (MDM) optical transmission system that uses deep learning neural network (DLNN) for Multiple Input Multiple Output (MIMO) detection is presented. Two channels operating at 250Mbps are QPSK modulated and transmitted at different mode through a Multi-Mode fiber and successfully detected. For MIMO detection, a supervised DLNN, which is designed, trained and evaluated using a Keras library and TensorFlow, is implemented in this MDM optical transmission system. The performance of our DLNN for MIMO detection is compared with Zero Forcing and Semi-Definite Relaxation Row-by-Row detectors. Our DLNN outruns the performance of these MIMO detectors.

引用

页数：6

共 5 条

[1]

Kingma D. P., 2015, P INT C LEARN REPR, P1

[2] Asynchronous OCDM-Based 10 G-PON Using Cascaded Multiport E/Ds to Suppress MAI Noise [J].