Robust adaptive beamforming using LMS algorithm with SMI initialization

The least mean squares (LMS) algorithm is a simple adaptive beamforming algorithm that is well suited for continuous transmission systems. The LMS algorithm converges slowly when compared with other complicated algorithms, such as recursive least squares (RLS) (Shubair, R.M. and Merri, A., 2005). On the other hand, the sample matrix inversion (SMI) algorithm has a fast convergence behavior. However, because its speedy convergence is achieved through the use of matrix inversion, the SMI algorithm is computationally intensive. Moreover, the SMI algorithm has a block adaptive approach for which it is required that the signal environment does not undergo significant change during the course of block acquisition. The paper develops an algorithm for robust adaptive beamforming by combining the attributes of the LMS algorithm and SMI algorithm. This new algorithm uses the LMS algorithm, which is simple to implement and not computationally intensive, but with SMI initialization in order to ensure fast convergence. Numerical results verify the improved convergence, accuracy, and computational efficiency of the combined LMS/SMI algorithm.