Study On The DOA Estimation For Coherent Wideband Signals Based On Matrix Reconstruction Technology

One of the basic problems on array signals processing is to estimate the Direction of Arrival (DOA) angle for wideband signals with array sensors. Current study about the DOA estimation for wideband signals has achieved plentiful fruit, but many methods have its own lack at certain extent. Based on the problem of the current methods and the goal about getting excellent performance of DOA estimation for wideband signals, a main study direction about the DOA estimation for wideband signals is to design new method model or optimize the conventional method frames.In these years, the study fruits about the DOA estimation for wideband signals mainly include incoherent signal subspace method (ISSM), coherent signal subspace method (CSSM), weighted average of signal subspaces (WAVES) and test of orthogonality of projected subspaces (TOPS). Firstly, ISSM applies narrowband techniques to get the DOA estimation based on every frequency bin, and then average the result from each frequency bin to get the final DOA estimation. Since the signal-to-noise ratio (SNR) varies at different frequency bin, so the performance of ISSM is not enough steady. Secondly, CSSM needs initial angles estimation to construct focusing matrix, so the performance of which is directly influenced by the initial estimation angle. In fact, poor initial values can lead to biased estimates and spoil the final estimation result. Thirdly, WAVES can avoid the initial-value requirement, but the beam-forming matrix depends on the size of a field of view and the array geometry. In addition, the WAVES will show worse performance if it construct focusing matrix with the same initial value as CSSM. Fourthly, TOPS does not require focusing angles or beam-forming matrix and can get excellent DOA estimation at larger SNR, but it has worse estimation performance compared with CSSM and WAVES at low SNR. To solve the above problem, this paper proposed three kinds of improved DOA estimation algorithms based on matrix reconstruction. The main work of this paper is shown as follows:1. This paper summarized the main algorithms about DOA estimation for narrowband signals. Mainly introduce some classic methods such as MUSIC, ESPRIT, SSF, fourth-order cumulants technologies, and give the corresponding program simulations. Moreover, the paper introduces the compressed sensing (CS) theory, and analyses its application on DOA estimation.2. This paper proposed a new DOA estimation algorithm for narrowband coherent signals independent to the signal subspace algorithms. Based on the influence from the array manifold at a hypothesized azimuth angle over the covariance matrix, an array manifold and a constant vector are used to modify the first line elements of the covariance matrix and a matrix with full rank is constructed. Then, a conclusion is proved that the rank of the reconstructed full rank matrix will increase when the angle of the array manifold is same with the real DOA of any one signal, contrarily the rank of the reconstructed matrix will keep unchanged as the original covariance matrix. Based on this conclusion, the DOA of the sources can be estimate with the reverse of the (P+1)th singular of the reconstructed matrix , and where the P is the number of signals. Simulation test proved the proposed algorithm is effective to estimate the DOA for narrowband coherent signals.3. This paper introduced the main DOA estimation algorithms for wideband signals. Mainly analyze the generation manners of the wideband signals, and generate the wideband signals with three kinds of manners. Moreover, the mathematics models about the array outputs of the wideband signals are introduced. Based on the above work, the method structures of the CSSM and TOPS are mainly introduced, and the corresponding performances of which are tested by the program simulation.4. This paper proposed two kinds of improved DOA estimation algorithms for wideband coherent signals. Firstly, this paper reconstructed a full rank matrix with no array aperture lack based on a narrowband correlated matrices coming from the DFT decomposition of the array outputs,then the conventional algorithms such as MUSIC are used to get the DOA estimation. This proposed algorithm has simple method structure and less computational cost compared with the conventional methods such as CSSM. Especially, the proposed method has excellent performance against the coherent wideband signals received by the sensors array. Secondly, this paper proposed an improved method based on TOPS method and its lacks of the performance. The proposed method solved the problem that the TOPS can not estimate the DOA for coherent wideband source signals at low signal noise ratio. The simulation showed that the proposed method has similar performance compared with RSS method. Furthermore, the improved TOPS inherits all of the virtues of the conventional TOPS, it doesn't require initial estimation about the DOA of the signals and has no independent focusing processing, so it has better performance compared with the RSS method in view of practicability.