分布式多传感器阵列的信号重构方法通信与信息系统专业论文.docx

哈尔滨 哈尔滨工业大学工学硕士学位论文 – – II – Abstract Traditional processing involves separate parameter estimation at individual sub- arrays and data processing at the fusion center. However, coherence between dif- ferent arrays are neglected, which can be utilized by joint coherent processing of distributed arrays. In this paper, we define basic concepts and derive geometrical and signal model on distributed arrays. Theoretical performance of coherent and noncoherent processing is analyzed with the help of error ellipse and Crame′ Rao Bound(CRB)0Parameter estimation can benefit from coherent processing. SNR, DOA, and source distance are factors that affect system performance. Array interpolation maps irregular distributed array to a well-behavior ULA, which can benefit from its Vandermonde structure. We considered effects of sector width, calibration direction number, and inter-distance of subarrays. Orientation of subarrays are vital to interpolation performance. When the virtual array share a similar direction with subarrays, performance are relatively better. Actually array interpolation is an uneasy task because no philosophy on virtual array selection is available. It is better to look for the Vandermonde structure from a mathematical view. The manifold separation technique (MST) models the steering vector of antenna arrays with arbitrary geometry, which can be expressed as prod- uct of a sampling matrix (dependent on the antenna array only) and a Vandermonde structured vector (dependent on the wavefield only). The calibration measurements are corrupted by noise in real-world applications, which impairs the performance of MST-based algorithms. The effect of noisy measurements on subspace-based DOA algorithms is analyzed. A link between the optimal number of selected modes and the statistics of noise is established. Simulation results show satisfactory performance. Keywords Distributed array; coherent processing; Crame′r-Rao lower bound; array interpolation; manifold se