Sparse And Tensor Representation Based Target Parameter Estimation In Radar-Communication Fusion System

Radar sensor networks(RSNs)are playing an important strategic role in national security and have penetrated into all aspects of society.Compared with traditional radar systems,RSNs are facing more challenges,such as shortage of spectrum resources,coexistence of distinct modules,miniaturization and energy consumption,which can render the deterioration of node and inter-module interference,and mobility limitation.The main cause of these problems is that the design method with separate modules cannot fulfill the requirements of increasingly dense RSNs.Against this background,the fusion design of communication and radar systems is envisioned to be a promising technology to tackle these challenges.As the differences between the communication and radar systems are gradually reduced,e.g.,the hardware,the working frequency and the signal characteristics,fusion is of great importance,especially for massive deployments of RSNs.Whereas,it is crucial and challenging that the implementation of joint signal processing.The challenges include how to allow the access of multiple users with limited spectrum resources,and how to achieve high detection performance with unaffected communication capacity.Meantime,traditional signal processing is sensitive to transmitted information and received noise.For this case,sparse representation has been showing evident advantages.Apart from this,with the widely applications of Multiple-Input Multiple Output(MIMO)technology,the performance improvements with higher dimensional signals are simultaneously highlighted.In this respect,the tensor representation,which can be viewed as the extension of the sparse representation in the higher dimensions,can effectively take advantage of the dimension extension in the signal processing processes.Inspired by the discussions above,this thesis focuses on the target parameter estimation problems based on sparse and tensor representation techniques,with the Orthogonal Frequency Division Multi-plexing(OFDM)based radar-communication fusion system.The main research achievements of this thesis include:The problem of target parameter estimation with tensor decomposition is first studied in this thesis.Most of the traditional signal processing methods in the fusion system are based on a vector or matrix model,which cannot take full advantage of the multi-dimensional structures of the received signals.The customary methods usually process all the dimensions sequentially,which may result in a loss of structural information and affect the target parameter estimation performance.On the other hand,sequential methods generally need additional parameter pairing procedure and may introduce ghost targets.In this thesis,target parameter methods with tensor decomposition are addressed.Firstly,the target estimation problem is modeled with tensor representation and the tensor decomposition algorithm is utilized to convert the multi-dimensional joint parameter estimation problem into multiple one-dimensional parameter estimation problems.For the case that the traditional tensor decomposition algorithm is unstable and is easily getting stuck in local extremum,a tensor power method with multiple random initializations is proposed to realize the optimal rank-1 tensor approximation.A greedy CP decomposition algorithm is proposed to solve the tensor decomposition problem when the rank is larger than 1.The proposed tensor power method and greedy CP decomposition algorithm improve the stability effectively and reduce the chance of getting stuck in local extremum.In order to utilize the parametric structural features of the factor vectors,a parameterized rectification algorithm is proposed.The simulation results demonstrate that the proposed algorithm improves the performance of target parameter estimation effectively.The target parameter estimation problem for the radar-communication fusion system with undersampling has been addressed in the second part of this thesis.In order to improve the accuracy of target parameter estimation,it is usually necessary to use a signal with large time-width bandwidth product.However,this not only imposes extremely high requirements on the hardware platform,but also causes waste on spectrum resources.Therefore,we propose a target parameter estimation method based on tensor completion when the transmitted and received signals are undersampled.Since the degrees of freedom in the target parameter estimation problem is much smaller than that of traditional tensor decomposition and tensor completion algorithms,we propose a functional tensor decomposition algorithm,in order to improve the target parameter estimation performance.Focused on the specific problem in this thesis,the optimal rank-1 functional tensor approximation is realized by a rough estimation on the coarse grids and subsequently refinements with Newton iterations.The greedy algorithm is used to solve the tensor completion problem when the rank is greater than 1.In order to reduce the computational complexity,a Fourier transform domain preprocessing method is proposed.The simulation results show that the proposed algorithms have much better performance than the existing algorithms in different undersampling methods and target scenarios,especially with low SNRs.In the last part of this thesis,the multi-user access scheme and corresponding signal processing method in the radar-communication fusion system are investigated.In the traditional communication systems,the block spectrum access and the uniformly-spaced spectrum access methods are usually adopted.However,these methods have great influences on the target parameter estimation performance.Moreover,the traditional signal processing method in the fusion system is not capable of coping the multi-user scenarios.Aiming at the multi-user access problem in the fusion system,we propose a multi-user access method based on cooperative random subcarrier allocation.The distribution characteristic of inter-user interference is analyzed for the proposed multi-user access method.Sparse representation based signal processing method is adapted at the receiving end.In order to reduce the computational complexity,a Doppler threshold detection method is proposed,which can effectively reduce the amount of computation required by the algorithm.The simulation analysis is carried out for scenarios without inter-user interference and with inter-user interference.The results show that the proposed method has much better performance than the traditional method in the fusion system.