Inter-pulse stagger radar signals are widely used in electronic countermeasures due to their anti-reconnaissance characteristics, and the correct classification of them can greatly improve the effectiveness of countermeasures. The primary and pivotal part of classification is the construction of effective features. However, there are few effective features are established based on the characteristic of inter-pulse stagger radar signal and suitable for their classification, which makes correct classification of such signals a difficult problem. Here, a novel feature extraction method of inter-pulse stagger radar signal is proposed in this paper. In this method, a modified permutation entropy analysis method which combines permutation entropy with improved variability measure is first utilized for encoding original stagger radar signal. And then, the complexity feature model of encoding result is established according to Lempel-Ziv algorithm and entropy. Finally, based on the extracted feature, Support Vector Machines is employed to classify different stagger radar signals to verify the feasibility and effectiveness of this complexity feature. The simulation experiment results show that all classification accuracies have reached over 97%. It is worth mentioning that the simulation shows that complexity feature proposed in this paper is not sensitive to a certain proportion of missing pulses and spurious pulses.