Low-Complexity Set-Membership Normalized LMS Algorithm for Sparse System Modeling
This work addresses computational efficiency in adaptive filtering for sparse systems, representing an incremental improvement over existing methods.
The authors tackled sparse system modeling by proposing two low-complexity set-membership normalized LMS algorithms (LCSM-NLMS1 and LCSM-NLMS2) that exploit system sparsity, achieving similar performance to state-of-the-art sparsity-aware algorithms while requiring lower computational cost.
In this work, we propose two low-complexity set-membership normalized least-mean-square (LCSM-NLMS1 and LCSM-NLMS2) algorithms to exploit the sparsity of an unknown system. For this purpose, in the LCSM-NLMS1 algorithm, we employ a function called the discard function to the adaptive coefficients in order to neglect the coefficients close to zero in the update process. Moreover, in the LCSM-NLMS2 algorithm, to decrease the overall number of computations needed even further, we substitute small coefficients with zero. Numerical results present similar performance of these algorithms when comparing them with some state-of-the-art sparsity-aware algorithms, whereas the proposed algorithms need lower computational cost.