An Incremental Recognition Method for MFR Working Modes Based on Deep Feature Extension in Dynamic Observation Scenarios

Multifunctional radar (MFR) working mode recognition is a crucial component of analyzing their behavioral intent, serving as a foundation for determining threat levels and making jamming decisions. Most of the existing methods for recognizing working modes are aimed at static observation scenes where all working mode classes are obtained at once. However, it is difficult to effectively adapt to dynamic incremental observation scenes. In this article, an incremental recognition method for MFR working modes based on deep feature extension (DFE) in dynamic observation scenarios is proposed. This algorithm can continuously update the knowledge of new classes while maintaining the performance of the learned classes through DFE and sample feature playback of old classes. The experimental results indicate that the proposed method can effectively meet the incremental recognition demands of the MFR working mode in a dynamic incremental observation scenario. Moreover, it performs well even under nonideal conditions, such as measurement errors, lost pulses (L-Ps), and spurious pulses (S-Ps).