Performance Research On LMS Adaptive Beamforming Algorithm

Adaptive beamforming is a significant technology of the smart antennas. Its core principle is that using adaptive algorithm to direct the main beam towards the desired signals and generate deep nulls towards the interfering signals. Thus, the desired signals'radiated power is maximized, while interfering signals'is minimized. The adaptive algorithm plays an important role in adjusting the weight vector. The adaptive beamforming algorithm, which determines the adaptive array system's performance, is a key part and research object of the adaptive array system.Faster convergence speed, lower computational complexity and more stable robustness is always the ambitions that researchers are pursuing. Around this issue, this paper mainly studies adaptive algorithms of the adaptive beamforming system. The paper's research work contains the following aspects:Firstly, the paper has introduced the developing process, application status and some key technologies of the smart antennas, explained the basic principles of the smart antennas and the adaptive beamforming, and built the basic model of the transmission signal in smart antennas. At the same time, various existing adaptive beamforming algorithms and their developing situation are synthesized in detail. Also, three fundamental adaptive beamforming criteria are summarized orderly, and every criterion is analysed and studied particularly and compared with each other. Three common adaptive beamforming algorithms are clearly elaborated: LMS algorithm, RLS algorithm, SMI algorithm. More simulation verifications are given to compare their performance. This paper is directed against the effect of the eigenvalue expansion to LMS algorithm convergence performance and the influence of different convergence factors to the misadjustment to open the discussion. At last of this part , the paper mainly presents a performance evaluation of the LMS adaptive beamforming by consider the effects of parameters including adaptive step size, the number of elements, the spacing between the elements and also the angular separation between two beams during the incident signals are received.Secondly, some major LMS-based algorithms are discussed, including NLMS algorithm, transform domain LMS algorithm and variable step size LMS algorithm. Moreover, this paper also pays much attention to the existing popular variable step size LMS algorithms. According to the various iterative manners of the variable step size in the weight vectors'iterative formula, the variable step size LMS algorithms could be distributed into three classes innovatively: the first class modifies the current time step through the last time step; the second class changes the step size value by using some functions of the square normalized form of error vector or input signals; the third class adjusts the step via some kinds of nonlinear functions. Furthermore, more simulation verification and comparison are provided to improve the modified algorithms of the three classes.In addition, an improved variable step size algorithm VSSDCTLMS is put forward. In view of the analysis of the basic transform domain DCT-LMS algorithm and the variable step size algorithm, the new algorithm is provided to overcome the problems of the traditional transform domain algorithm and variable step size algorithm, such as the slow convergence speed and the sensitivity to the correlated signals. The new improved algorithm has faster convergence speed than the traditional transform domain algorithm at the early convergence stage, while keeps low misadjustment by using small step size values at the steady state. Compared to the traditional variable step size algorithm, the new algorithm provides higher anti-interference. Moreover, the new modified algorithm shows excellent tracking capability when the system is in the case of mutation, and keeps very fast re-convergence speed. In the meantime, the properties of the new algorithm are simulated verifiably carefully. At last the new algorithm is applied to the adaptive beamforming model.At last, the thesis fulfills the function simulation of LMS-type adaptive beamforming based on FPGA. The design, whose advantages is fast operation, small consumption of resources and easy realization, realizes the function simulation of LMS adaptive beamforming by the premise of guaranteeing the precision of operation and Effective cut-off bits of data. The simulation results show that the design achieves the convergence of the LMS algorithm effectively, and at the same time displays the convergence curve of the weight vectors accord with the theoretical analysis.