Research On Interference Recognition And Anti-interference Technology In Airspace Based On Neural Network

As an important part of the electronic countermeasure system,jamming pattern recognition is the premise of the implementation of targeted anti-jamming measures.At the same time,the recognition of all kinds of new electronic jamming or even compound jamming is the focus of radar communication countermeasure research.After the interference identification is completed,anti-jamming technologies such as time-domain,airspace and space-time combination should be adopted to ensure the normal operation of radar system.Neural network has a wide range of applications in pattern recognition and solving convex optimization problems.Aiming at the shortcomings of the traditional interference recognition method based on neural network,which has complex steps and the recognition accuracy is greatly affected by human factors,an automatic interference recognition method for deep learning radar based on long and short term memory unit(LSTM)is proposed.Considering that the beamforming technology can enhance the desired signal and suppress the interference and noise,a new robust adaptive beamforming algorithm model is constructed based on the existing optimization algorithm,and an efficient and stable feedback neural network algorithm is used to solve the optimization problem to obtain the optimal weight.The main work of this thesis will focus on the above aspects.Firstly,this thesis proposes a radar interference recognition method based on LSTM network,which can automatically learn the internal distinguishing features of different kinds of interference signals,so the network only needs to input the time-frequency sequence information of interference types to automatically complete the task of classification and recognition.This thesis first introduces the structure diagram of the cyclic neural network and the LSTM network,and explains that the neural network model is suitable for processing the long-time sequence signal data.Then,the interference time-frequency sequence information is input into the network for training.During the training process,the super parameters are constantly adjusted to achieve the best network structure.The trained network is used to identify unknown interference types.The simulation results show that compared with the traditional interference recognition method based on PNN,the proposed method does not need to preprocess the signal and extract the eigenvalues,and has a higher accuracy for interference type recognition.In the part of robust beamforming technology,based on the worst performance optimization algorithm and norm constrained optimization algorithm,a new robust adaptive beamforming algorithm is proposed.This algorithm uses inequality constraints in the optimization cost function,reduces the freedom consumption of the array system,and effectively improves the interference suppression ability of the system.Theoretical analysis shows that the proposed algorithm can improve the robustness of the error of the steering vector and covariance matrix by restricting the sensitivity of the beamformer.Theoretical analysis also shows that the algorithm can be classified as diagonal loading algorithm.Finally,a feedback neural network structure is used to solve the optimization problem and get the optimal weight.It can be proved that the proposed algorithm has less sensitivity of beamformer than the optimization algorithm under the worst performance,so it has better robustness in the presence of steering vector error and covariance matrix prediction error.The simulation results show that the proposed algorithm has performance advantages over the classical algorithm in the presence of multiple interferences,low number of snapshots and multiple steering vector errors.