Research On Target Tracking Algorithm For Distributed Sources

In most applications of array signal processing, conventional target tracking techniques are generally based on point source modeling, where the energy of each source is assumed to be concentrated at discrete direction angles that are referred to as the source direction of arrival. However, in applications such as wireless communication, radar, sonar, and etc, the transmitted signal is often obstructed by buildings, vehicles, trees, and etc. The phenomena of multi-path depending on the scattering may result in angular spreading of the source energy. In this case, a distributed source model, the high-order signal model will be more appropriate. Therefore, the target tracking in distributed source environment has been a popular topic in array signal processing.In this thesis, we analyze the cause of distributed source and its characteristic, and particularize the model of distributed source. Considering the fast target tracking problem of time-varying central direction of arrival for distributed source, this thesis has studied several classical central direction of arrival subspace tracking algorithms, including projection approximation subspace tracking algorithm, orthogonal projection approximation subspace tracking algorithm, approximated power iteration subspace tracking algorithms and fast approximated power iteration subspace tracking algorithms, and then analyze and compare the advantages and disadvantages of the studied algorithms. For practical application problems, an efficient method for the central direction of arrival tracking algorithm based on the support vector regression is proposed. The proposed algorithm provides a fast and effective solution for the central direction of arrival tracking in the case of small samples.To combine the array signal processing with the support vector machine, the proposed algorithm is based on the use of a support vector regression approach for the approximation of the unknown mapping that performs the transformation from the outputs of the elements of the smart array to the angles of arrival. Using the array received covariance matrix as input, and the central direction of arrival of the distributed source as the output characteristics, I train the support vector regression model, using the indirect approximation method estimating the central direction of arrival to achieve the purpose of tracking. The approach reduces the approximation error caused by the output discontinuity in the case of direct approximation. Simulation results show that the proposed algorithm can effectively solve the issue of the central direction of arrival tracking of distributed source, and in the case of larger angular spread the proposed algorithm has strong adaptability.