Parameter Estimation of LFM Signals Based on LVD in Complicated Noise Environments

In view of reducing the effects of cross terms, conventional methods of parameter estimation for Linear Frequency Modulation (LFM) signals suffer from low accuracy and huge computational complexity. To solve these problems, LVs Distribution (LVD) based method is introdused in this paper. It provides directly accurate Centroid Frequency-Chirp Rate (CFCR) representation of a LFM signal. The rescaling operator is used for the Parametric Symmetric Instantaneous Autocorrelation Function (PSIAF) to eliminate the effects of linear frequency migration on the time axis, then a two-dimensional (2-D) Fourier transform is taken over the new scaled time variables to convert a 1-D LFM signal into a 2-D single-frequency signal. The resulting signal can be represented with distinct peaks on the CFCR plane, whereas the energy of the cross terms can be ignored compared with the peaks of auto terms. The coordinate values of LFM components directly correspond to their centroid frequency and chirp rate. LVD can suppress effectively the Gaussian noise, however, the performance of the CFCR domain analysis for signals in heavy-tailed impulsive noise environment is in severe degradation. Considering this issue, an improved Fractional Lower Order LVD (FLOLVD) for the stable distribution noise is proposed. Computer simulation results show that the proposed approach obtains high-accuracy phase estimation, and it is robust to the impulse noise as well as the Gaussian noise.