Study Of High Resolution DOA Estimation Algorithm And Its System Implement

Parameter estimation is an important research direction of modern signal processing, and has been widely applied in the field of radar, sonar, biomedical, communications, and thus the parameter estimation method has an important theoretical and practical values.In this paper, we study the parameter estimationmethod of DOA, frequency and distance under the Gaussian white noise, themain work are summarized as follows:1. We follow the thoughts of classical DOA estimation methods and thenpropose a method for joint2-D angle and carrier frequencie estimation.This method rearranges the L-shaped placed sensors into a vector. Thekey idea of the proposed method is the usage of the time-domain rotationproperty under the assumption of the narrow-band signals, and theneatimate the carrier frequency and the arraymanifold. At last, thematching property of the resultant carrier feruency and arraymanifold isused to estimate the2-D angle. The proposed method has higherresolution in distingush closely-located signals either in azimuth orelevetion angle. It is notable that the proposed method is different fromthose troditional2-D estimation methods whose estimation only use thespatial information, the proposed method exploits frequencies to increasethe information used for localizing the far-field signals. As a result, theproposed method has a better performance in estimating DOA comparedwith the troditional methods. In the end, numerical experiments are madeto illustrate the efficiency of the proposed method.2. We improve two available near-field localization methods and propose afast algorithm that does not need any search procedure. The formermethod needs spectrum search either in angle estimation or in range estimation. Then the latter method improves the former with a close-formresolution of angles while the ranges remain no improvements. Intackling this problem, we propose a close form resolution of ranges. Theproposed method is based on the property of rotational invariance, onwhich the relationship between two subarray is obtained. Then we exploitthe property of rank deficiency of two matrices to propose aRoot-MUSIC-based algorithm. This proposed method is comparable tothe former method in performance while the complexity is much lessthan the two existing methods. Numerical experiments and comparasionswith the former method are made and the results show the effectivity ofthe proposed method.3. We propose a fast algorithm for three-dimensional single near-fieldsource localization with uniform circular array. The proposed method hasclose form resolution. The key idea of this algorithm is to construct twocorrelation functions by using the centrosymmetric property of UCA. Thetwo correlation functions contain information of azimuth/elevation angleand range of the impinging signal. From the constructed two correlationfunctions we exploit LS method to estimate the3-D parameters.Computation al complexity of the proposed method is also analyzed.Results of the numerical experiments show that the proposed LS methodis comparable to the3-D MUSIC method in performance while thecomputational complexity is much less than the3-D MUSIC.