CCSAE-Based Un-Cooperative Communication Behavior Recognition Scheme

Abstract

The task of non-cooperative communication behavior recognition (CBR) usually faces a complex electromagnetic environment, and the interfered monitoring data will affect the accuracy of communication behavior recognition. A convolutional conditional staked auto-encoder (CCSAE) based un-cooperative communication behavior recognition scheme is proposed in this paper. In particular, the proposed CCSAE denoising module can filter out the noise caused by complex electromagnetic interference by adding conditional constraint to the auto-encoder (AE) structure, and the deep convolutional AE structure can better extracts high-dimensional features related to communication behavior. By comparative experiments, it can be found that the CCSAE-based CBR scheme can stably and effectively improve the accuracy of un-cooperative communication behavior recognition task under complex electromagnetic environment.