基于FRFT的雷达信号chirp基稀疏特征提取 及分选

 特征分析是雷达信号分选识别的基础,利用稀疏分解思想对新体制雷达信号进行特征提取是一个新的研究方向。本文以分数阶Fourier变换的核函数作为稀疏分解的chirp基函数,将具有相近特征参数的chirp基函数构成基函数族用于稀疏分量提取,推导了在分数阶Fourier域基于匹配跟踪的chirp基函数族稀疏分解公式,然后利用chirp基稀疏分量的调频率和初始频率构成特征参数序列,将雷达信号脉冲分成5大类进行分选和识别,仿真分析验证了推导结果的有效性。结果表明对于具有线性或曲线时频特征的雷达信号在信噪比为-3 dB,采样频率为500 MHz,观测时间为2 μs,调频率不超过100 MHz/μs时,仍然具有95%的正确分选概率。

Chirp Function Sparse Feature Extraction and Sorting of Radar Signals Based on FRFT

Feature analysis is the basis of radar signal sorting and identifying, and feature extraction of a new system radar signals through sparse decomposition is a new research topic. This paper uses the fractional Fourier transform kernel function as the basic chirp function of sparse decomposition to make up chirp functions with similar parameters into a function family for extracting sparse components, and derive the sparse decomposition formula with a matching pursuit in the fractional Fourier domain. Then a characteristic parameter sequence is formed consisting of chirp-based sparse components’ chirp-ratio and initial frequency, and the radar signal pulses are divided into five classes for sorting and identifying. Simulation analysis proves the validity of the derived conclusion, and results show that the linear or curved time-frequency characteristics of radar signals still have 95% correct sorting probability when the SNR is -3 dB, sampling frequency is 500 MHz, observed time is 2 μs, and the chirp-ratio is no more than 100 MHz/μs.