基于小波能量谱和ReliefF算法的雷达辐射源无意调制特征提取

Feature extraction of unintentional modulation of radar emitter based on wavelet energy spectrum and ReliefF algorithm

为了获取更加有效的雷达辐射源无意调制特征并进一步降低特征的维度,以提升低信噪比下雷达辐射源个体识别的准确率,从时频分析角度出发提出了一种基于小波变换能量谱和ReliefF算法的无意调制特征提取方法。首先对辐射源信号进行小波变换并获取小波能量谱,然后采用ReliefF算法对小波能量谱值进行权重分析,筛选出区分能力较强的高权重小波能量信息作为雷达辐射源的无意调制特征。该方法将权重分析应用于特征提取中,在提升特征有效性的同时进一步降低了特征的维度。实验结果表明:相较于传统时域和频域中的无意调制特征,基于小波能量谱和ReliefF算法提取的无意调制特征具有低维度、强抗噪声的特点。当信噪比大于0 dBm时识别率达到90%以上,具有较高的工程应用价值。

In order to obtain more effective unintentional modulation feature of radar emitter and further reduce the dimension of the feature,so as to improve the accuracy of radar emitter recognition under low SNR.In this paper,a method of feature extraction of unintentional modulation based on wavelet energy spectrum and ReliefF algorithm is proposed from the perspective of time-frequency analysis.Firstly,wavelet transform is applied to the emitter signal to obtain the wavelet energy spectrum,then the ReleifF algorithm is used to analyze the weight of wavelet energy spectrum value,and the high-weight wavelet energy information with strong discrimination ability is selected as the unintentional modulation feature of radar emitter.In this method,weight analysis is applied to feature extraction,which improves the effectiveness of the feature and further reduces dimension of the feature.The experimental results show that,compared with the traditional unintentional modulation features in traditional time domain and frequency domain,the unintentional modulation features extracted based on wavelet energy spectrum and ReliefF algorithm have low-dimension and strong anti-noise characteristics.When the signal-to-noise ratio is greater than 0 dBm,the recognition rate reaches more than 90%,and has high engineering application value.