http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
오학준,정찬수 崇實大學校 生産技術硏究所 2001 論文集 Vol.31 No.-
This paper compares and analyzes the performance of the ECLMS algorithm and conventional LMS algorithms under a double-talk situation. The LMS algorithm has been a popular method to train adaptive FIR filters for echo cancellation due to its simplicity. In a double-talk situation, the LMS algorithm performs poorly when both near-end and far-end signals are presented. The error signal for obtaining the gradient becomes considerably large in such a double-talk situation. Conventional LMS algorithms usually stop adaptation when this situation happens to keep the coefficients freeze under a double-talk condition. Stopping rap adaptation is just a passive action to handle double-talk conditions, which results in lower adaptation speed. To solve this problem, expanded correlation LMS (ECLMS) algorithm has been proposed. It utilizes correlation as an input signal instead of the input signal itself. This algorithm could adapt the parameters continuously even in the double-talk situation, and showed good convergence property compared with conventional LMS algorithms as the normalized LMS (NLMS).