Research Article
Deep Learning Based Single-Channel Blind Separation of Co-frequency Modulated Signals
@INPROCEEDINGS{10.1007/978-3-030-41114-5_45, author={Chen Chen and Zhufei Lu and Zhe Guo and Feng Yang and Lianghui Ding}, title={Deep Learning Based Single-Channel Blind Separation of Co-frequency Modulated Signals}, proceedings={Communications and Networking. 14th EAI International Conference, ChinaCom 2019, Shanghai, China, November 29 -- December 1, 2019, Proceedings, Part I}, proceedings_a={CHINACOM}, year={2020}, month={2}, keywords={Single-channel blind separation (SCBS) Deep learning (DL) Bidirectional recurrent neural network (BRNN)}, doi={10.1007/978-3-030-41114-5_45} }- Chen Chen
Zhufei Lu
Zhe Guo
Feng Yang
Lianghui Ding
Year: 2020
Deep Learning Based Single-Channel Blind Separation of Co-frequency Modulated Signals
CHINACOM
Springer
DOI: 10.1007/978-3-030-41114-5_45
Abstract
This paper presents our results in deep learning (DL) based single-channel blind separation (SCBS). Here, we propose a bidirectional recurrent neural network (BRNN) based separation method which can recover information bits directly from co-frequency modulated signals after end-to-end learning. Aiming at the real-time processing, a strategy of block processing is proposed, solving high error rate at the beginning and end of each block of data. Compared with the conventional PSP method, the proposed DL separation method achieves better BER performance in linear case and nonlinear distortion case with lower computational complexity. Simulation results further demonstrate the generalization ability and robustness of the proposed approach in terms of mismatching amplitude ratios.
