Research On Robust Sparse Reconstruction Models And Algorithms In DOA Estimation

Direction-of-arrival DOA estimation is one of two main research directions in array signal processing, and hence has a wide range of applications in radar, sonar, and communication. A large number of nonparametric and parametric approaches have been proposed for DOA estimation in the past half century. Recently, a class of semiparametric algorithms based on sparse representation has attracted significant interest due to its capability in achieving high resolution and dealing with coherent sources.Most of the above techniques are developed based on the assumption of exactly known array manifold. However, in practical applications, sensor position errors are inevitable due to the circumstance and so on. Performance degradation or even failure will occur in the presence of sensor position errors. Therefore, DOA estimation in the presence of sensor position errors is an issue with great significance and practical value.DOA estimation with sparse representation models based on the vectorization of the covariance matrix in the presence of sensor position errors are mainly studied in this thesis. For narrow-band far-field signals, from the perspectives of the vectorization of the covariance matrix of array data, a new robust sparse representation model and DOA estimation algorithm based on this model is proposed. The main contributions and innovative points in this thesis are listed as follows:1. For the issue of sensor position errors, based on the assumption of the location perturbations obeying normal distributions, a new sparse representation(SR) model is formulated based on the vectorization of the covariance matrix. The statistical characteristics of perturbations are considered and no Taylor approximation errors are introduced in the SR models. Simulation results show that compared with existing robust DOA estimation methods and self-calibration DOA estimation methods, the robust sparse reconstruction model and method proposed in this paper have better performance.2. For the evaluation of DOA estimation performance, we have derived the CramerRao lower bound(CRLB) of four signal models. Simulations and experimental studies have been carried out. In particular, the parameters of the sensor position error statistics,signal to noise ratio, the snapshot number and spatial signal incident direction affect the CRLB have been analysed. The results provide a theoretical basis for research and analysis of DOA estimation in real environment.