Full-duplex fast initializing digital subscriber loop echocancellers

摘要

In this paper, we propose a novel approach to fast initializingndigital subscriber loop (DLS) echo cancellers using full-duplexntransmitted training sequences. The important difference between ournmethod and the previous half-duplex fast training schemes is that we cannsimultaneously estimate the impulse responses of the echo paths of bothnends rather than estimate them separately. Therefore, the initializationntime is greatly reduced. The effects of the noise and symbol rate offsetnbetween two ends are examined in terms of both mean-square error (MSE)nand signal to noise ratio (SNR), and the results show that the newnalgorithm is robust in the presence of noise and symbol ratenoffset