A stochastic model for the transform-domain LMS algorithm

This paper presents a stochastic model for the transform-domain LMS (TDLMS) algorithm. Here, we derive a new stochastic model to describe the first and second moments of the adaptive filter weights. The proposed model takes into account the time-varying nature of the normalized step-size parameter arisen from the power normalization stage, as well as it disregards the use of auxiliary expressions. Comparing with Monte Carlo simulations, we verify the accuracy of the proposed theoretical models.