Frequency-hopping Signal Radio Sorting Based on Stacked Auto-encoder Subtle Feature Extraction
Ping Sui, Ying Guo, Hongguang Li, Shaobo Wang, Xin Yang · 2019
The prior information of the enemy frequency-hopping (FH) radio stations that can be obtained in real time is limited, under the non-cooperative conditions of the battlefield. In the conventional radio signal-sorting algorithm based on parameter estimation of FH signals, the number of estimation parameters that can be used for sorting is small, the estimation accuracy is low, and the accuracy of sorting largely depends on the estimation precision of parameter, therefore the effects of radio signal sorting cannot be realized in a complex battlefield environment. In order to solve the above issue and learn from the idea of signal subtle feature extraction of neural network, this paper proposes an FH signal radio-sorting algorithm based on subtle feature extraction of stacked auto-encoder. The method utilizes the stacked auto-encoder to extract radio feature, then realizing the FH signal radio sorting by Low-rank clustering algorithm. The experimental results of the measured FH signals demonstrate the feasibility of the proposed method.