- 作者: 陳巽璋
- 作者服務機構: 國立中山大學電機系
- 中文摘要: 快速最小均方值適應性濾波演算法則,是牛頓法的一種近似法。一般認為它的收斂特性較其他種演算法則為佳。同時對於輸人信號的特性變化較具軔性。本算論文之目的是想以線增益器為分析對象就其暫態及穩態之收歛特性,作理論分析。由分析中我們發現這種演算法的收斂速度不隨輸人信號正弦波信號間能量分佈而改變。而只與演算法之步距有關。這與傳統最小均方值適應性濾波演算法則不同。證明本演算法則實有較佳之收歛特性與軔性。
- 英文摘要: The rapid complex least mean square (LMS) adaptive algorithm is an approximated form of theNewton method used in the adaptive filtering problem. Its convergence rate, in general, is considered to befaster and more robust than the existing algorithms used in adaptive filtering applications. That is,the algorithm is less sensitive to variation in the input signal in the adaptation process. In this paper, a statistical analysis of such an algorithm is carried out for further investigation of thetransient convergence property. To obtain an explicit expression, in terms of mean square error (MSE),the adaptive line enhancer (ALE) is considered. From computer simulation, we show that the convergencerate of the rapid algorithm in the application of ALE depends only on the values of step-size. It does notdepend on the disparity in eigenvalues of the input autocorrelation matrix. However, this is not the casewith the conventional complex LMS adaptive algorithm having the same input signal statistics.
- 中文關鍵字: rapid convergence adaptive algorithm; transient behavior; eigenvalue spread; lineenhancer
- 英文關鍵字: --