A wavelet based partial update fast LMS/Newton algorithm

Abstract

This paper studies a wavelet based partial update fast LMS/Newton algorithm. Different from the conventional fast LMS/Newton algorithm, the proposed algorithm first uses a shorter-order, partial Haar transform-based NLMS adaptive filter to estimate the peak position of the long, sparse channel impulse response, and then employs the fast LMS/Newton algorithm integrated with partial update technique to fulfil the rest convergence task. The experimental results demonstrate the proposed algorithm outperforms its conventional counterpart in convergence performance and possesses a significantly lower computational complexity.