Adaptive beamforming via two-dimensional cosine transform

A novel transform domain least-mean-square (LMS) algorithm for weight adaptation in the Griffiths-Jim generalized sidelobe canceler is presented. In this algorithm the set of tap-input vectors at the output of the blocking matrix is considered as a 2-D array and the whole array is mapped to transform domain via a full 2-D orthogonal transform. A self-orthogonalizing LMS algorithm is then used for weight adaptation in the transform domain. The purpose of the 2-D orthogonal transform is to remove both the autocorrelation of the signal within individual channels and the cross-correlation between adjacent channels. As a result the proposed algorithm exhibits faster convergence rate than previous algorithms and is therefore well-suited for the real-time processing of nonstationary array signals