Radar Waveform Recognition Based on Multiple Autocorrelation Images

Abstract

Radar signal waveform recognition, as a key component of radar target recognition, has always been a research topic of great concern in the field of electronic countermeasures. In this paper, aiming at the contradiction between improving recognition rate and reducing the sample size, we propose a multiple autocorrelation joint decision models, which achieves higher waveform recognition rate while requires lower data volume for the original sampled data. The key point of the model is to perform multiple autocorrelations on the signal and use time-frequency transform for each times autocorrelation result to obtain multiple time-frequency feature images that can characterize the same signal. Then, the model adapts to the input of multiple feature images and gets the pre-classification results of each feature image. Finally, using the pre-classification results, this paper designs an inference machine module based on a fully connected structure to achieve a better signal classification result. This paper simulates six types of the signals and generates training sets and test sets, using two data sets to achieve hyper parameter optimization, training, and testing of the model. The simulation results compared with the literature show that the proposed model not only has a high recognition rate at a high signal-to-noise ratio (SNR) but also better adapts to waveform recognition at low SNR environment. At -9dB SNR, the recognition rate of six types of signals is more than 74%.

Keywords

• Autocorrelation
• inference machine
• low SNR
• radar signal waveform recognition