In order to extract robust features for radar high-resolution range profile (HRRP) recognition, a feature extraction method using circular convolution coefficients of HRRP is proposed in this paper. The features extracted by the proposed method represent the HRRP structure information which is more robust compared with local details. Moreover, this feature can make up for the loss of HRRP local details caused by feature selection. Due to the high dimensionality and nonlinear separability of HRRP features, feature selection based on kernel class separability was also used. Experiment results on simulation HRRP data sets are analyzed to demonstrate the efficiency of our method.
展开▼
