摘要
针对实际侦测辐射源目标信号交叠、参数灵活多变、波形不完整等特点,以及传统目标分类识别方法准确率不高的问题,提出了一种基于深度特征的信号分类方法。通过数据积累对全脉冲数据进行补全、纠错,综合时域、频域、空域、能量域、调制域等多域参数提取隐藏特征,搭建深度学习网络对侦测数据进行持续训练,实现对复杂目标信号的分类。
In view of the characteristics of overlapping target signals,flexible parameters and incomplete waveforms in actual detection of radiation sources as well as the low accuracy of conventional target classification and recognition methods,a signal classification method is proposed based on depth features.The full pulse data are complemented and error corrected through data accumulation,and obscure features are extracted from multi-domain parameters such as time domain,frequency domain,space domain,energy domain,and modulation domain.The deep learning network is built to continuously train the detection data to achieve the classification of complex target signals.
作者
臧勤
洪鼎
钱鸥
刘佳媛
尚睿
ZANG Qin;HONG Ding;QIAN Ou;LIU Jia-yuan;SHANG Rui(No.8 Research Academy of CSSC, Nanjing 211153)
出处
《雷达与对抗》
2021年第1期43-45,53,共4页
Radar & ECM
关键词
深度学习
隐藏特征
目标分类
deep learning
obscure features
target classification