Entropy feature extraction approach for radar emitter signals

Abstract

Radar emitter signal recognition is an important and diffrcult issue in electronic intelligence, electronic Support measure and radar warning receiver systems and is always Traditional methods cannot recognize advanced radar emitters in modern electronic warfare. So aiming at the Characteristics of approach (EFEA) is proposed in this paper. The main points of [3i and non-ambiflity phase EFEA are that an improved approximate entropy (ApEn) restoral approach [41, were Presented to analyze different method and norm entropy (NE) method are presented to extract radar emitter signals. The methods have some drawbacks in features from radar emitter signals. ApEn can measure engineering applications because quantitative analysis is not quantitatively the complexity and irregularity of radar emitter made and varying signal-to-noise rate (SNR) is not signals from a relatively small amount Of data. NE iS a measure considered. Therefore, the methods cannot be applied to of uncertainty, imbalance and disorderliness of signals. EFEA has good characteristics of easy implementation and short computation time. In the experiment, 9 typical radar emitter signals are chosen to make an experiment of feature extraction main'y in and signal recognition. Experimental results demonstrate that processing. Entropy-based describe average accurate recognition rate amounts to 98.28% in a large information-related properties for an accurate representation range of signal-to-noise rate because ApEn feature and NE of a given signal. [51 Approximate entropy (ApEn) is a feature have good characteristics of clustering the same signals statistic that can be used as a measure to quantify the and separating the different signals and have strong stability, complexity and irregularity of both deterministic and which indicates EFEA is effective and practical. stochastic signals. ApEn is firstly presented by Pincus [6] to evaluate the complexity and irregularity of complex systems. short computation time and ApEn can discem changing complexity and irregularity from a relatively small amount of