Cognitive Radar Signal Processing Based on Convolutional Neural Network

Single-criterion is difficult to meet the radar’s multi-working mode and multi-tasks, a signal processing method using convolutional neural network (CNN) for cognitive radar is proposed, which uses the learning ability of CNN to learn the nonlinear relationship between one-dimensional data to joint the mutual information (MI) criterion and signal to noise ratio (SNR) criterion. Firstly, a train set composed of signals generated by MI and SNR criteria and the corresponding environmental information was constructed; Secondly, designing a CNN model based on the form of data set; Using the local connection and global sharing of CNN to reduce the number of network parameters and improve training efficiency. Using CNN to efficiently extract features of one-dimensional data from MI and SNR criteria to generate signal features, which are sent to the fully connected layer for regression to achieve the purpose of combining MI and SNR criteria. Finally, the CNN model is trained on the train set and test by test set. Simulation results prove that the proposed method can joint the MI criterion and SNR criterion, and improve the comprehensive performance of radar.