基于矩特征的雷达信号脉内调制样式识别方法研究

摘要

In order to obtain low probability interception (LPI) and high distance and speed resolutions, phased array radar signals require a complex intra-pulse modulation mode so that the transmission energy can spread to a wide frequency range, sometimes the transmission signal is close to the noise. Through recognizing the intra-pulse modulation mode of radar signals, we can achieve enemy radar's operational performance and formulate the corresponding strategy of interference. Aiming at the characteristic that different modulation signals have different geometric distributions of time-frequency images, the Choi-Williams transform is used for five types of signals, which are CW, LFM, PARFM, PSK and FSK signals. Four invariant moments that reflect the geometric features of time-frequency images with different SNR are presented as the input vectors, the gradient descent algorithm is adopted to build a neural network, and then the modulation modes are recognized. The simulation recognition results verify the effectiveness of the proposed method.%相控阵雷达为了获得良好的低截获(LPI)特性,同时兼顾距离和速度分辨率,需要对信号进行复杂的脉内调制,将发射能量扩散到很宽的频率范围上,甚至与噪声水平相近.对雷达脉内调制信号进行识别,可以获取敌方雷达作战性能并制定相应的干扰策略.针对不同调制信号的时频图几何分布不同的特点,对CW、LFM、PARFM、PSK、FSK五类信号进行Choi-Williams变换,提出将不同信噪比下时频图像中最能反映图像区域几何特征的四个不变矩作为输入特征向量,利用梯度下降算法构建神经网络,对调制样式进行识别的方法.仿真结果表明,方法的识别效果良好.