基于熵特征的雷达辐射源信号识别

Radar emitter signal recognition based on entropy features

针对现有方法识别率低和没有考虑噪声影响的问题,提出一种新的雷达辐射源信号识别方法。将近似熵(ApEn)和范数熵(NoEn)构成特征向量,用神经网络分类器实现自动分类识别。ApEn是定量描述信号复杂性和不规则性的有效测度,NoEn是定量表征信号能量分布的有效参数。理论分析和实验结果表明,熵特征类内聚集性强、类间分离度大,在较大信噪比范围内均能获得非常满意的正确识别率,证实了所提出方法的有效性。

To solve the problems of low recognition rate and noise effect in radar emitter signal recognition, a novel approach was proposed. In this approach, approximate entropy （ApEn） and norm entropy （NoEn） constituted feature vector, and neural network based classifiers were designed to identify radar emitter signals automatically. ApEn is a good measure of complexity and irregularity of signals and NoEn is a useful parameter for quantifying the energy distribution of signals. Theoretical analysis and experimental results show that ApEn and NoEn features have small within-class distance and large between-class distance, and can achieve very satisfying accurate recognition rate when signal-to-noise rate varies in a large range. It is proved to be a valid and practical approach.