基于BDS-GD的低截获概率雷达信号识别

LPI radar signal recognition based on BDS-GD

针对在信号特征提取与识别中使用双谱估计数据量大、维度高的问题,提出了双谱对角切片(BDS)与广义维数(GD)相结合的识别方法。通过提取信号双谱对角切片减少数据量,并利用多重分形理论中的广义维数降低数据维度,对切片内部特性进行细微描述,基于距离测度提出特征评价指标,从而选出最具有区分度的3个阶数对应的广义维数作为特征向量,输入到最小二乘支持向量机中进行分类识别。使用4种低截获概率(LPI)雷达信号作为待识别信号,仿真结果表明,本文方法提取的信号特征在特征空间中有良好的聚集性和离散性,在0 dB信噪比下,识别准确率能达到92.2%,与选取的其他方法对比说明其具有很好的识别性能。

Regarding the deficiencies of bispectrum with big data and high dimension in signal feature extraction and recognition,a method combining bispectra diagonal slice（BDS）with generalized dimension（GD）was proposed.First,BDS was used to reduce data volume and GD in the multi-fractal theory was taken to reduce dimension in order to make subtle description for slice.Second,generalized dimension corresponding to three ranks is treated as feature vectors by feature evaluation index based on distance measure.Finally,the feature vectors will be input into least squares support vector machine for recognition.Four sorts of low probability of intercept（LPI）signal are used to be recognized,and the simulation results show that the signal features in feature space have good aggregation and discreteness,and the accuracy rate of recognition can reach 92.2% when the SNR is 0 dB,which shows that it has good performance in recognition compared with other algorithms.