Variable Step-Size Discrete Cosine Transform Diffusion LMS over Multi-task Networks

摘要

In this paper, new variable step-size transform domain (VSSTD) algorithm is developed for Diffusion Least Mean Square (DLMS) multi-task networks with system identification application over Wireless Sensor Networks (WSNs). Our main contributions are the theoretical derivations of convergence analysis of Discrete Cosine Transform DLMS (DCTDLMS) algorithm considering adaptive combiners. Our simulations showed performance improvement compared to the traditional DLMS.