Research On Realization Of DOA Estimation Algorithm Based On Fast Eigenvalue Decomposition

DOA which is a fundamental problem in array signal processing has a very important application in the field of military and civilian use such as radar,sonar,communication,seismic survey and radio astronomy.The basic problem of DOA estimation is to determine the spatial position of multiple attentive signals existing in a region of space.In the past four decades,the field of the estimation of the DOA has sprung up a number of high-resolution algorithms.Among them,the MUSIC algorithm proposed by R.O.Schmidt et al in 1979 has been widely used because of its high resolution,high estimation accuracy and high stability in some certain situations.However,the electronic reconnaissance and countermeasures has a high demand at the real-time array signal processing and MUSIC algorithm operation process is complicated on account of the decomposition of matrices which has a large amount of computation and a large number of nonlinear operation,so the rapid implementation of MUSIC algorithm has become a major callenge for the current array signal processing.In the recent years,FPGA technology develops rapidly which has been laid a solid foundation of the rapid implementation of MUSIC algorithm.This article is based on the project of Direction of Arrival Estimation of Narrowbaed Frequency Modulated Signals with Missing Time Samples in which the implementation of a real-time direction finding system of 4 element uniform linear array based on FPGA is a key link.The specific work of this article are as follows:1.For the problem of excessive number of iterations in the CORDIC calculation unit in the EVD module,the dissertation introduces the MAR CORDIC algorithm into the EVD module calculation.Meanwhile the dissertation proposes a solution to solving the problem that MAR CORDIC algorithm's expansion factor is not precarious when modular correction is on.And the dissertation uses the Verilog language for RTL level description to completing the software simulation of MAR CORDIC algorithm and the traditional CORDIC algorithm.2.Based on the parallel characteristics of Jacobi algorithm,the improved EVD module based on BLV array structure and MAR CORDIC algorithm was studied and the main diagonal processing unit,the non-main diagonal processing unit and the eigenvector unit was designed.The matrix EVD module software simulation was completed by using theVerilog language for RTL level description and the EVD module computing speed was enhanced.3.The covariance module and the peak search module were described by using Verilog language at RTL level,and the software simulation is completed.4.The three modules of the MUSIC algorithm were connected and debugged,and the implement of the MUSIC algorithm at RTL level based on the fixed point is completed.The design of the entire project is based on the Xilinx Virtex-5 series XC5VSX95 T and ISE 14.4 software development system.The calculation time of 4?4 matrix's EVD in the 110 MHz maximum clock frequency is less than 1 ?s and accuracy of EVD has been ensured.The computing time of EVD is shorter than the most current domestic and foreign project.The implementation time of MUSIC algorithm is 6.5 ?s which is decreased by three orders of magnitude than that of DSP-based MUSIC dedicated processor.The implementation time of MUSIC algorithmto is also faster than the current type of implementation.