- 作者: 陳巽璋; 陳柏義
- 作者服務機構: 國立中山大學電機系
- 中文摘要: 在本篇論文中,我們提出一種混合式最小均方值濾波器之演算法則。這個演算法則是擷取傳統的最小均方值演算法則與轉換空間最小均方值演算法則之優點來完成所需要之收斂特性。 在此,我們探討了兩種類型的混合式最小均方值演算法則,即固定步距與可調變步距。並且利用計算機模擬在通訊系統頻道等量化的應用上來探討它的收斂特性。 一般而言,由在暫態時的收斂速度與穩態時的均方值誤差來說,混合式最小均方值演算法則比傳統及轉換空間的方法較俱軔性。除此之外,在可調變步距演算法則中,對於如何利用中間值濾波器來選擇步距大小我們也提出了一些意見
- 英文摘要: Adaptive filtering techniques have been intensively used in various engineering applications. Inthis paper, we are concerned with devising a new adaptive filtering algorithm, named the hybrid least-meansquare (HLMS) algorithm. The concept behind the HLMS algorithm is to use the merits of the convention-al least-mean square (LMS) and the transformed domain LMS (TDLMS) algorithm to achieve the desiredperformance. Two types of the HLMS algorithm, viz., the constant step-size and the variable step-size, areexamined. The convergence properties of these two HLMS algorithms are investigated using computersimulations with application to the channel eqalization of communication systems. In general, the HLMS adaptation algorithm performs more robustly than the conventional LMSalgorithm or the TDLMS algorithm in terms of the convergence rate in the transient state and the steady-state mean-square error (MSE). Furthermore, a selection of the criteria for reducing the step-size in thevariable step-size algorithm using the median filtering technique is suggested.
- 中文關鍵字: hyrbid; discrete cosine transform; adaptive filter; convergence rate; median filter; channel equalizer
- 英文關鍵字: --