Study On Electronic Reconnaissance Technology Based On Sparse Reconstruction

Electronic reconnaissance, which is the foundation of fight, command and strategic dicision, is an important field in electronic countermeasures. It can collect intelligence in peacetime and provide support in wartime. In modern complex electromagnetic environment, traditional analog reconnaissance receiver can hardly meet the modern requirements. With the development electronic reconnaissance technology, digital reconnaissance receiver will take more important roles in the future. As the performance of digital reconnaissance receiver depends on the signal processing algorithm, it has become an important research topics to explore effective algorithms in electronic reconnaissance. In recent years, the sparse reconstruction and compressed sensing theory become the focus of research in the field of signal processing. It can reconstruct signal by a small amount of observed data based on the signal sparsity. To improve intercept probability, the reconnaissance receiver tends to cover the whole spectrum and space, and the received signal over a period of time is sparse in frequency domain or spatial domain. Therefore, applying the sparse reconstruction and compressed sensing theory to electronic reconnaissance has a broad development prospect.This dissertation studies the electronic reconnaissance technology based on sparse reconstruction, including high precision frequency estimation, high resolution array direction-finding, joint frequency and angle estimation, and electronic reconnaissance technology based on signal intercepted samples, etc. The relevant work is supported by the National Basic Research Program of China(973 Program) and the National High Technology Research and Development Program of China(863 Program). The main content of this dissertation is summarized as follows.The first part focuses on the frequency estimation problems. The channelized frequency measuring method based on polyphase filter and subspace decomposition frequency estimation method is analyzed and it is verified by simulations. And on this basis, a frequency estimation method for narrow-band signals based on sparse reconstruction is proposed. The redundant dictionary is established by pre-estimating frequencies, which reduces the dictionary length and computation effectively. The pre-estimating results are applied to the iterative process to seek the optimal matching atoms and update the residual vector. The simulations show that this method can achieve high estimation precision and has robust performance in low SNR cases.The second part focuses on the DOA estimation problems. A DOA estimation algorithm without knowledge of source number is proposed, which can estimate the DOA of multiple sources using fewer array elements by the array aperture extension. The simulations show that the algorithm has a higher estimation precision, resolution and can work in low SNR cases. The rationality for applying sparse reconstruction to DOA estimation is analyzed and the model is established, which is verified by simulations. Utilizing the energy-concentrated characteristic, a DOA estimation algorithm for LFM signal based on sparse reconstruction in FRF domain is proposed. By establishing the redundant dictionary respectively, it reduces the dictionary length and computation.The third part focuses on the joint frequency and angle estimation problems. A novel algorithm based on sparse reconstruction is proposed. The sparse model is established in spatial frequency domain and the pair matching for frequency and azimuth is achieved by spatial filtering and FFT. By extending it to joint frequency and 2-D angle estimation, a broadband 2-D DOA algorithm is proposed. By defining the spatial frequency of azimuth and elevation, it uses separable dimension processing and utilizes the signal frequency measured by FFT and the correlation of spatial filtering output data to achieve matching process. For limited array cases, a broadband DOA estimation based on nested arrays is proposed. It reduce the number of array and improve the precision and resolution by array aperture extension. Compared to spatial filtering, the pair matching is implemented by sparse reconstruction, which ad voids the matrix inversion and tends to be more simple and efficient.The fourth part focuses on the reconnaissance processing method for the known radar signal. By using signal intercepted samples to establish the redundant dictionary, the detection and recognition process is converted to the optimal matching atoms problem. The influence of initial phase, time synchronization error and pulse arrival time is analyzed. Compared to the traditional reconnaissance process, there is no need of parameter measurement process in this method, which is suitable for rapid detection and recognition of the specific threat radar signal.