DOA Estimation For Acoustic Vector-sensor Array

Vector sensor array is an array system with new system and new array signal processing. Combining it and modern signal processing technology will bring new impetus for array processing, and it has important significance for new system of sonar system and radar system development. Since it is posed, measuring system based on vector sensors and the vector sensor array have been get in research and application more and more, and signal DOA estimation algorithms also constantly emerge. These algorithms mainly base on the parametric and the spatial spectral.This paper aims at DOAs estimation methods for acoustic vector-sensor array based on pre-existing algorithm theory with higher precision and better performance. The main work can be summarized as follows:This thesis expounds the fundamental theory of the array signal processing systematically, and summarizes the developments, applications of acoustic vector-sensor array.An analysis on several array models and establish data model.The problem of the DOAs estimation is addressed with linear acoustic vector-sensor array, including ESPRIT algorithm and multi-invariance MUSIC (MI-MUSIC) algorithm. The former constitutes two same translation subarrays, using the signal rotation invariant to estimate the DOA of the acoustic vector-sensor array; The latter estimates the DOAs using only a one-dimensional (1-D) search, and it is able to enforce the constraint that the subarray responses for a given source are related by a scalar multiplier that lies on the unit circle. In addition, we compare their root mean squared error (RMSE) and costs of computational complexity against those of ESPRIT. The MI-MUSIC algorithm can have better DOA estimation performance than ESPRIT algorithm.The next section links the acoustic vector-sensor array DOAs estimation problem to the arbitrarily space and unknown location. First, we illustrate ESPRIT algorithm, which use the relationship with rotating factors of each subarray and orientation cosine of the target signal to estimate the DOA of the arbitrarily spaced acoustic vector-sensor array at unknown location. Then, we link the acoustic vector-sensor array parameter estimation problem to the trilinear model. Relying on the uniqueness of trilinear decomposition, a trilinear decomposition algorithm has been proposed to estimate two-dimension DOA for acoustic vector-sensor array. Our work extends the trilinear decomposition method to an arbitrarily spaced array, and our algorithm provides the DOAs estimation for irregularly or randomly spaced array manifolds, whose locations need not be known. This algorithm has better DOAs estimation performance than ESPRIT algorithm. In contrast to the classic subspace algorithm, this algorithm requires no a priori knowledge of the array geometry or peaking searching, and it does not require pair matching. Furthermore, our algorithm can support small sampling size.