The FLS-QR algorithm for adaptive filtering: The case of multichannel signals (Q809049)

The Fast Least Squares (FLS) algorithms based on the QR triangular decomposition of the input signal matrix and developed for the case of one-dimensional signals can be extended to handle the case of multidimensional (MD) or multichannel signals. The FLS-QR algorithms for multidimensional signals are conceptually more complicated than those for one-dimensional signals. However, they can be implemented using a simplified approach which is just a repetition of the 1-D algorithm for a K times run, for K-dimensional signals. In that case, the computational complexity is just multiplied by the number of input signals and the basic properties of the 1-D case should be kept, particularly the robustness to wordlength limitations.