Radar signal deinterleaving method exploiting correlation of multi-parameter time series

Signal deinterleaving is a critical technique in the process of radar electronic reconnaissance. However, the performance of the traditional signal deinterleaving algorithm is degraded dramatically when faced with high pulse loss ratios and noise pulse interference. To deal with this deficiency, a radar signal deinterleaving method exploiting the correlation of multi-parameter time series is proposed. TOA, RF, and PW are used to construct a pulse feature representation. The bidirectional recurrent neural network is used to explore the long-term temporal patterns in the pulse stream and extract the characteristics of the pulse time series context via supervised learning. Simulated experimental results showed that the proposed method could obtain robust performance under non-ideal conditions, which can achieve an accuracy of over 90% even if the pulse loss ratio reached 70%.