Logo image
在加成性白色高斯通道上應用於符號時序回復之遞迴式最大可能估測法的收斂性分析
Thesis

在加成性白色高斯通道上應用於符號時序回復之遞迴式最大可能估測法的收斂性分析

王文山
Masters, National Tsing Hua University
1995

Abstract

遞迴式最大可能估測法 一致性 漸近有效 同步 Recursive Maximum Likelihood Estimation consistent asymptotically efficient synchronization
此篇論文主要是探討在加成性白色高斯雜訊通道(AWGN)上估測脈衝振幅調變( PAM)訊號之符號時序的問題。解決這個問題的方法在傳統上大都基於極大化一個可能函數的準則,如追蹤迴路或最大可能估測法。但這兩種方法可能會遇到兩個不好的暫態追蹤形為。例如,當估測值很靠近正確值時,但更新步伐大小仍然很大,或當估測值離正確值還有一段距離時,但更新步伐大小已經變得很小時,這樣長的收斂時間會導致一個不是很有效的同步策略。文獻[26]提出了一個適用於在AWGN通道中作時序回復的遞迴式最大可能估測(RMLE)演算法。但到目前為止,尚未有人對RMLE作詳細的收斂性分析。在本論文中,我們探討RMLE的漸近式行為,包括機率為壹的收斂性和漸近分怖。RMLE的收斂問題可以回歸到一個相關於RMLE之非隨機常微分方程式固定點的穩定性問題。一旦證明了RMLE收斂的機率是壹,我們可以利用一些引理、簡單的運算和中央極限定理得到RMLE估測值序列的漸近分佈是高斯分佈且其漸近的變易量達到Cramer-Rao之下界值。更進一步,我們證明了RMLE演算法是一致性(consistent)且漸近有效(as ymptotically efficient),i.e.符號時序估測值之序列收斂到正確值的機率是壹(with probability 1)且漸近的誤差平方平均值達到Cramer-R ao之下界值。The problem of estimating symbol timing for PAM signals overadditive white Gaussian noise (AWGN) channels is considered.Conventional approaches to this problem are usually based onmaximization of likelihood functions. In \cite{Won-Lu}, arecursive maximum likelihood estimation (RMLE) algorithm hasbeen proposed suitable for timing recovery over AWGN channels.However, details on convergence analysis of the RMLE algorithmhas not been touched yet. In this thesis, we investigate theasymptotic behavior of the RMLE algorithm, includingconvergence with probability one (w.p.1) and asymptoticdistribution. The convergence problem of the RMLE algorithm canbe effectively reduced to the stability problem of stationarypoints of an associated deterministic ordinary differentialequation (ODE). Furthermore, we show that the RMLE algorithm isconsistent and asymptotically efficient.

Metrics

1 Record Views

Details

Logo image