Research on Individual Identification Method of Radar Radiation Source Based on Multi-layer and Multi-dimensional Input

Aiming at the problems of slow speed and low accuracy of radar radiation source individual recognition under traditional methods, this paper proposes and establishes a radar radiation source individual recognition method based on multi-level and multi-dimensional input. Specifically, a radar signal multi-domain feature extraction framework based on contrast predictive coding (CPC), wavelet transform analysis and statistical feature extraction was established to extract and mix the timing features, time-frequency features and statistical features of the target signal respectively. A 64 × 64 × 314 feature vector representing the radar signal is obtained, which is input to the ResNet residual neural network for training and individual recognition. At the same time, for the feature extraction method based only on CPC framework and the multi-domain and multi-channel hybrid feature extraction method, we further carried out comparative analysis experiments under different radar signal data. In the actual experimental verification, the average recognition rate of type 11 radar data is more than 97%, and the recognition effect is better than other feature extraction methods under the same conditions, which further verifies the application value of the model in practical engineering.