上海交通大学智能微波光波融合创新中心（imLic）硕士生易思成的工作——A multi-band low-noise microwave signal receiving system with photonic frequency down-conversion and transfer-learning network（基于光子下变频和迁移学习的多波段低噪声信号接收系统）的相关成果近期被Optics Letters期刊接收发表，该工作得到了国家重点研发计划（2019YFB2203700）、国家自然科学基金（61822508）的部分资助。该工作提出一种基于光子下变频和迁移学习的多波段低噪声信号接收系统。其中光子下变频系统直接接收微波信号，迁移学习网络用于降低信号中的噪声。除了有效的去噪能力外，迁移学习网络还具有针对不同类型或不同频带频率信号的超快速再训练功能。实验结果表明，该微波信号接收系统可以提高不同类型、不同信噪比、不同占空比接收信号信噪比8dB左右。同时，对于网络再训练，迁移学习网络只需要传统训练方法1/3的数据和1/10的时间消耗就能完成网络训练。
摘要: In this Letter, we propose and demonstrate a multi-band signal-receiving system, powered by photonic frequency down-conversion and transfer learning. A photonic frequency down-conversion system directly receives the microwave signals, and the transfer-learning network (TLN) lowers the noise in the signals. In addition to the effectiveness of denoising, the TLN also features ultra-fast retraining for signals of different types or different multi-band frequencies. Experimental results showed that the proposed microwave-signal-receiving system can improve the signal-to-noise (SNR) ratio of signals of different types, SNR, and duty cycles. For network retraining, the TLN requires only three times less data and 10 times less time consumption than conventional training methods.