A Multifunction Radar Work Mode Recognition Method Based on the LSTCN Network

The recognition of Multifunction Radar (MFR) operating modes in complex electromagnetic environments is a critical aspect of modern electronic warfare and countermeasures. Multifunction radars exhibit high flexibility and agile parameter adjustment capabilities, significantly increasing the complexity of the electromagnetic battlefield and posing substantial challenges for reconnaissance in electronic warfare. Traditional mode recognition methods struggle to effectively capture long- and short-term dependencies when dealing with complex time-series signals. To address this issue, this paper proposes an LSTCN model that integrates Long Short-Term Memory (LSTM) Network, Temporal Convolutional Networks (TCN) and the selfattention mechanism to recognize multifunction radar operating modes. The model is designed to extract long-term dependencies from time-series signals while utilizing TCN to capture short-term dependency features through dilated convolutions, thereby effectively modeling local temporal relationships. A dataset was generated based on the operating principles of multifunction radars and divided into training and testing sets to simulate operating modes under various environmental conditions. Experimental results demonstrate that the LSTCN model exhibits superior recognition accuracy and robustness under various interference conditions, outperforming current SOTA algorithms such as LSTM, GRU and TCN.