脉内特征提取是新体制雷达辐射源信号分选的关键问题，文中针对现有方法分选准确率不高和对噪声敏感的问题，提出了一种基于高次频谱相像系数和频域奇异谱熵特征的分选新方法，实现了低信噪比下雷达辐射源信号的高准确率分选。对接收到的信号提取高次频谱相像系数特征以及奇异谱熵特征，并将两者作为分选的联合特征向量，运用K means聚类算法实现对不同调制方式的雷达辐射源信号的分选。仿真结果表明：改进后提取的信号特征类间的分离度大且受噪声影响程度小，在信噪比为-2 dB的情况下，该算法的总体平均分选准确率在85％左右，不同调制类型信号间的分选准确率最低为80％。与现有方法相比，文中提出的算法具有更好的信号识别效果。

Intra-pulse feature extraction is a key issue in advanced radar emitter signal recognition. Aiming at the problem of low sorting rate and sensitivity to the signal noise of common methods, a new sorting approach based on the resemblance coefficient of higher order spectrum and singular spectrum entropy in frequency domain was proposed. High sorting rate of radar emitter signals is gotten under low signal to noise ratio(SNR). The resemblance coefficient of higher order spectrum and singular spectrum entropy in frequency domain of the received signal are extracted and they are used as unite sorting characteristics. The sorting of radar emitter signals with different modulation modes is completed by K means algorithm. The simulation results show that the features of signals after pretreatment have large between class separation degree and they are not sensitive to noise. The overall average classification accuracy rate of the proposed algorithm is about 85％, and the lowest signal sorting rate of different modulation modes is 80 when SNR is -2 dB. Compared with the existing approaches, the proposed approach has a better recognition effect.

TN957

上海市自然科学基金