DOA Estimation Algorithms For Noncircular Signals Using Sensor Arrays

A signal is called to be noncircular if its statistics are rotationally variant. The BPSK and AM modulated signals, which are widely used in communication systems, are noncircular. Enhancing the performance of the direction-finding algorithms by utilising the noncircularity of the signals is a prosperous area in the array signal processing domain. The existent direction finding algorithms for noncircular signals, whose major computation comes from eigenvalue-decomposition or singular value decomposition of the covariance martix of the array outputs, can hardly handle large number of signals impinging on the array with limited sensors. And the 2D-NC-UESPRIT algorithm, which can be applied to rectangular arrays and regular-hexagonal shaped arrays, is the only 2-dimensional direction finding algorithms for noncircular signals. In this thesis, several algorithms for noncircular signals are proposed to overcome the above knots. The main research work and contributions are detailed as follows.1. The performance of MUSIC algorithm for noncircular signals (NC-MUSIC) in the presence of modeling errors is analysed and the general flow of the performance analysis of the peak-searching algorithms is also presented.2. The MUSIC algorithm for noncircular signals based on even-order (2q) cumulants (NC-2q-MUSIC) is presented for the case of a number of signals impinging on the array with limited sensors and its performance in the presense of modeling errors is also analysed. The estimation capacity, the resolution, the angle-estimation precision and the robustness to modeling errors of the proposed algorithms all exceed those of conventional MUSIC algorithms with the same order. And all the above four aspects of the performance increase with q, which is validated by theoretical analyses and simulation results.3. The extended propagator method (EPM) for noncircular signals is proposed to alleviate the heavy computation of the eigenvalue-decomposition or singular value decomposition and its asymptotic performance is analysed. The EPM and NC-MUSIC share the same asymptotic performance. The performance of EPM in the case of low SNR and small number of snapshots is worse than that of NC-MUSIC. Fortunately, it can be overcome by over-estimating the number of signals properly. The real-valued version of EPM is proposed and its computational load is only a quarter of that of complex EPM.4. To fill the lack of 2-dimensional (2-D) direction finding algorithms for noncircular signals, the extended rank reduction algorithm for 2-D direction finding of noncircular signals (2D-NC-RARE) is presented based on two parallel linear arrays and the parameter weighted method for noncircular signals (NC-PWM) is also presented based on the L-shaped array. The asymptotic performance of the two algorithms is analysed. Profiting from the configuration of the arrays, the two algorithms need only 1-D search for joint azimuth and elevation estimation. Comparing with the existent 2-D direction finding algorithm for noncircular signals, i.e., 2D-NC-UESPRIT, the proposed algorithms are superior in estimation capacity and angle-estimation precision.