A Time-invariant Sparse Model and a Deep Unrolling Network for Target Detection of Passive Radar
In recent years,the target detection method based on sparse feature extraction has become a research hotspot in radar field.However,due to the uncontrolled transmitted waveform of orthogonal frequency division multiplexing(OFDM)-based passive radar,the sparse model will change with the unknown transmitted waveform,resulting in a large amount of calculation and more manual intervention for the corresponding target detection method.On the other hand,it is difficult to detect target echo because it is often covered by strong clutter such as direct-path signal.In this context,a time-invariant sparse model is proposed by using the waveform characteristic of the OFDM-based passive radar and the channel frequency response at the pilot position.Then,a realization method of intelligent passive radar target detection based on the deep unrolling network is firstly studied by replacing each iteration process of the sparse model solution with a layer of neural network.Simulated and measured results show that the proposed method has similar performance to the traditional clutter suppression method in target detection,but it has lower computational complexity,and does not need to manually design the solving parameters such as sparse matrix of sparse model.
passive radarorthogonal frequency division multiplexing waveformtarget detectionsparse modeldeep learning
NETL
NSTL
万方数据
