Real-time detection of high-frequency RF signals requires sophisticated hardware with large bandwidth and high sampling rates. Existing microwave photonic methods have enabled sub-Nyquist sampling for bandwidth-efficient RF signal detection but fall short in single-shot reconstruction. Here we report a novel single-shot sub-Nyquist RF signal detection method based on a trained deep neural network. In a proof-of-concept demonstration, our system successfully reconstructs high frequency multi-toned RF signals from 5x down-sampled singleshot measurements by utilizing a deep convolutional neural network. The presented approach is a powerful digital accelerator to existing hardware detectors to significantly enhance the detection capability.
|Title of host publication||MWP 2018 - 2018 International Topical Meeting on Microwave Photonics|
|Publisher||Institute of Electrical and Electronics Engineers Inc.|
|Publication status||Published - 03 Dec 2018|
|Event||2018 International Topical Meeting on Microwave Photonics, MWP 2018 - Toulouse, France|
Duration: 22 Oct 2018 → 25 Oct 2018
|Name||MWP 2018 - 2018 International Topical Meeting on Microwave Photonics|
|Conference||2018 International Topical Meeting on Microwave Photonics, MWP 2018|
|Period||22/10/2018 → 25/10/2018|
Bibliographical noteFunding Information:
This work was supported in part by the EU FP7 Marie-Curie Career Integration Grant (631883), in part by the Royal Society (IE170007), in part by National Natural Science Foundation of China (Projects 61771148, 61571211 and U1501251), and in part by Guangzhou Science and Technology Plan (Project 201607010290).
© 2018 IEEE.
Copyright 2019 Elsevier B.V., All rights reserved.
- Convolutional neural network
- deep learning
- Nyquist sampling
- under sampling
ASJC Scopus subject areas
- Computer Networks and Communications
- Signal Processing
- Atomic and Molecular Physics, and Optics