Radar emitter signal identification using intra-pulse parameters plays an essential role in radar countermeasure systems. Current research has been conducted on mathematical transformation domains such as the time domain, the frequency domain, or the time-frequency domain. However, no matter how to improve the classifier model, there still exists a lower accuracy rate of deinterleaving or recognition under a complex modern radar environment. Thus, this paper proposes a feature extraction method using complex networks theory for recognition of radar emitter signals. Firstly, a dynamic segmentation model based on the BIC information criterion is introduced to construct the complex network from the signal frequency series. Secondly, the statistical dynamic characteristics which reflect topological indices of different complex networks are extracted separately. Finally, these features after principal component analysis are taken as input to the KNN model for clustering. The experimental results show that the proposed method of features extraction achieves a recognition rate of over 92% in eight typical radar emitter signals at Signal-Noise Ratios of 0dB, 10dB, 20dB.