Research On Parameter Estimation Algorithm For Distributed Sources

Conventional direction-of-arrival (DOA) estimation algorithms are generally based on the assumption that objective is point information source. There exists sight transmission in the dissemination process. But in real environments, instead of direct wave, there exist rich scattering and multipath effects, which made the traditional DOA estimation performance seriously deteriorated. Because distributed source signal model introduces more parameters which can better describe the actual signal propagation characteristics, it has a wide range of applications in mobile positioning system, low elevation radar tracking system, sound localization system and sonar systems. The study of parameter estimation algorithm under distributed source environment has attracted wide attention from scholars.This paper begins with the discuss of background of distributed source, researching status of distributed source and key issues, and then introduce the distributed source signal model and its characteristic in detail, study the basic algorithms of distributed source parameter estimation deeply. Aimed at the small sample problem in practical applications, an algorithm that based on support vector regression (SVR) for distributed source parameter estimation is proposed. This algorithm provides an effective solution for parameter estimation of distributed sources in the case of effective range of the sample and low signal-to-noise ratio (SNR).The support vector regression that base on statistical theory is applied to array signal processing techniques. The regression parameters are selected on the base of particle swarm optimization (PSO) algorithm. The vector after being handled its autocorrelation matrix of the sample and the direction-of-arrival estimation of distributed sources as training samples, together with the regression function that has been trained are used to estimate the central DOA of distributed sources. The process of obtaining parameter estimation values is a process of solving the regression function values. Compared with the traditional algorithm, the proposed algorithm reduces the computational complexity and is easy to implement. Simulation results show that the proposed algorithm can more effectively solve the distributed source central DOA estimation of distributed source under the circumstances of effective range of the sample and low signal-to-noise ratio.