Research On Methods For Joint DOA And Frequency Estimation Based On State-Space Model

In the research field of signal processing, array signal processing plays an important role. Its main research directions include adaptive array signal processing and space spectrum estimation (or direction of arrival (DOA) estimation). The past thirty years saw fast development of the direction of arrival estimation technology, while most research was focused on improving the estimation accuracy of multiple parameters of spatial signals, and its applications including sonar, radar, mobile communications, radio astronomy, exploration, and biomedical imaging etc.Recently, the joint estimation of DOA and frequency becomes a hot research topic in the field of array processing. When the model parameter method was deployed during signal processing, polynomial models like AR, MA and ARMA were widely used. On the other hand, the parameter estimation method based on state space model has attracted lots of attention. By taking SVD to the state space matrix, it can achieve better estimation accuracy and noise attenuation performance. This thesis addressed the problem of joint estimation of DOA and frequency based on the state space model, where the signal is assumed to be far-field narrow band with Gaussian colored noise. The research work can be summarized as follows:1. For uniform linear array, a new joint estimation method was proposed based on the subspace analysis algorithm. This method built the state space model, and transferred the estimated parameter to the system matrix in the state space model. The system matrix was estimated from another matrix, and then DOA and frequency were derived from that system matrix. The effectiveness of the algorithm was simulated.2. For uniform linear array, a joint DOA and frequency estimation method was proposed based on fourth-order cumulant. This method constructed the fourth-order cumulant matrix for the output signal, and the pre-estimated value of DOA and frequency were derived, by iterating the state space model, and taking SVD to the state matrix etc. The simulation proved the effectiveness of the algorithm.3. For the L-type linear array, joint DOA and frequency estimation method was created based on the fourth-order cumulant. This method used L-type linear array to process the algorithm in different dimensions, transferred a three dimension estimation into a one dimension and two dimension estimation, and all parameters were obtained by only taking one SVD computation. The correctness of the algorithm was simulated.Compared to other methods, the proposed method in this thesis has following characteristics:1. The state space model provides some level of flexibility for model parameterization. The estimation accuracy is high and noise attenuation capability is good as well.2. The subspace method directly distinguish the system state space model, it does not require parameterization and iterating. The implementation of the algorithm only relies on SVD calculation, suitable for distinguishing multiple valuable systems.3. High-order cumulant can depress any Gaussian noise (white or color), it can expend array aperture, and has no requirements on array's geometrical shape. Moreover, the computation of the noise covariance matrix is not needed.4. The methods in this thesis do not require complex search of spectrum peak, the DOA and frequency will be automatically paired, with suitable computation complexity.