Specific Emitter Identification Based on Wavelet Convolution Supervised Contrast Learning

Specific emitter identification technology has an important role in military and civilian fields, and it is a new idea to apply contrast learning in the field of specific emitter identification. The traditional supervised contrast learning algorithm has the problem of low recognition accuracy, so we propose a wavelet convolution supervised contrast learning ( Wavelet Convolution Supervised Contrast Learning, WC-SupCon) algorithm for specific emitter identification. Firstly, the bispectral maps of I/Q signals are generated; then, they are inputted into the feature extraction layer of WC-ResNet (Wavelet Convolution ResNet), and the parameters of the feature extraction network are updated and trained by the supervised contrast loss function; finally, the parameter weights of the feature extraction network are kept constant, and the cross-entropy loss function is used to train the classification network, which can complete the task of specific emitter identification. specific emitter identification task. The experimental results show that the accuracy of specific emitter identification based on wavelet convolution supervised contrast learning is 92.47%, which is 12.59% higher than that of the traditional network. Therefore, the method proposed in this paper has more practical value for contrast learning in the field of specific emitter identification.