Noise-robust radar HRR target recognition based on complex Gaussian statistical models

Abstract

In order to utilize the phase information to improve radar target recognition performance of high range resolution (HRR) target signals, this paper takes adaptive Gaussian classifier (AGC) model and factor analysis (FA) model for instance to generalize Gaussian statistical models to complex domain to build complex HRR models. It is demonstrated that the structures and parameter estimators of complex Gaussian statistical models are invariant to the initial phases of HRR complex data. Furthermore, to enhance the recognition performance under low signal-to-noise ratio (SNR) conditions, a noise-robust modification algorithm is introduced. It solves the mismatch between complex Gaussian models and noisy test signals. Experimental results show that the proposed models can obtain higher average correct recognition rates by utilizing the phase information. Also, the modified models can deal well with noisy test signals.