Higher-order Statistics-based Doa Estimation Techniques

A novel class of analysis tool in signal processing-Higher-Order Statistics (HOS) is presented in this thesis. We then use it along with the well-developed algorithms such as ESPRIT and MUSIC in direction-of-arrival (DOA) estimation.First, the thesis introduces the definitions and the attributes of the Higher-Order Statistics. It is insensitive to additive Gaussian noise (white or colored), which is what we base on to DOA problems. Then two DOA estimation algorithms based on Higher-Order Statistics are presented, one is that forming cumulant matrix pencil used in ESPRIT to estimate DOA problems, the other is spectrum estimation method for DOA estimation based on the eigenstructure analysis of the fourth-order cumulant, and comparing the effects of the estimation to conventional covariance-based DOA algorithms'. Second, since there are lots of redundant items in the fourth-order cumulant matrices, we introduce an algorithm of direction-of-arrival estimation based on the Maximal Set of Nonredundant Cumulant (MSNC). The algorithm takes advantages of all nonredundant information of the fourth-order cumulant matrices and so reduces the computational cost. Comparing to other conventional DOA algorithms, simulation experiments illustrate the potential performances of the proposed algorithm. Third, a novel class of fourth-order cumulant matrices for DOA estimation is presented. Both the order and the components of the cumulant matrices can be adjusted suitably to improve the performance.At last, the method to advance the performance of the estimation in limited data samples is studied. We carry out the computer simulations comparison and analysis to the algorithms presented above.