基于BP神经网络的残缺LFM信号识别方法

Recognition of Incomplete LFM Signal Based on BP Neural Network

在雷达对抗接收机截获雷达信号时,由于失谐、带宽不匹配等因素导致信号部分丢失,针对传统特征提取算法难以有效提取出信号特征,导致残缺信号识别正确率低的问题。以线性调频信号为例,采用基于Adam优化算法的BP神经网络模型,对不同丢失率条件下的带噪线性调频信号进行识别,分析噪声和数据丢失率对识别精度的影响。实验结果表明,信噪比大于-3dB,数据丢失率小于30%时,可以达到92%以上识别率。

In the background of partial signal loss caused by detuning and bandwidth mismatch when the radar countermeasure receiver receives the signal,the traditional feature extraction al-gorithms are difficult to extract the signal features effectively,which result in the low recognition accuracy of the recognition of the incomplete signal.Linear frequency modulation(LFM)signal is taken as an example and BP neural network based on Adam optimization algorithm is adopted,LFM signals with noise under different loss rates are identified,and the influence of noise and data loss rate on the recognition accuracy is analyzed.The experimental results show that the recognition accuracy rate can be more than 92%when the signal-to-noise(SNR)is greater than-3dB and the data loss rate is less than 30%.