An Intra-Pulse Modulation Type Recognition Algorithm for Radar Signals Based on the Improved Residual Network

Abstract For purpose of solve the problem poor discrimination and robustness intra-pulse signal features extracted by traditional methods, this paper proposes a radar modulation type recognition algorithm based on improved residual network. Firstly, one-dimensional time-domain is converted into two-dimensional time-frequency image Smoothing Pseudo Wigner-Ville Distribution; Then preprocessed; ResNet-50 network chosen as framework. In order to retain feature map information much possible, convolution kernel increased in module. The cross entropy loss function center are used speed up convergence realize signal. simulation experiments show that when SNR -14dB, overall average accuracy for eight kinds signals (CM, LFM, NLFM, BLFM, BPSK, QPSK, OPSK, LFM+BPSK) can reach 97.29%, which shows effectiveness.