ORTHOGONAL LMS ALGORITHMS FOR FAST LINE ECHO CANCELER TRAINING

摘要

An echo canceler for calculating an echo impulse response during a training sequence, wherein a sequence of data samples is selected such that a full rank data correlation matrix comprised of such data samples is diagonal at the Nth iteration. The echo canceler estimates the echo impulse response based on such data samples and an echo generated during the training sequence using a normalized least means square algorithm, wherein the length of the data sequence to sound the echo channel is N. This normalized LMS algorithm, wherein the data sequence is specifically selected, converges toward an echo impulse response much faster than standard LMS algorithms, while having a computational complexity comparable to that of the standard LMS algorithm.