可適性水聲通道追蹤之最佳遺忘因子估測

Abstract

水聲通訊和無線電通訊主要有二個最大的差異點，一是水聲通道有非常長的多重路徑延遲，範圍可涵蓋十到一百多個符號(symbols)，另一個是通道時變的速度。對於基於通道估測的等化器來說，通道估測是決定其效能的表現的最重要因素。在本篇論文中以遞迴性最小平方法做於通道追蹤的演算法，在這種演算法中若使用固定的遺忘因子去追蹤時變通道是不適合的，因為若當遺忘因子接近1時收斂會變慢，但當遺忘因子過小時又有錯誤調整過大的問題。所以水聲環境的通道追蹤必需可適性的調整遺傳因子以得到較好的追蹤效能，論文中我們提出一種較為直接的方法，以一種實現理論最佳值的方法來設定遺傳因子，取代以往以殘差均方誤差做於依據來調整遺傳因子的方法。最後實驗的效能表現是以「ASIAEX」的資料做為測試來源，在實驗中會以「信號多徑比」(SMR)做為衡量通道狀況的好壞。在實驗結果中，可以觀察到我們提出的遺傳因子估測法非常的有效，即時在較為嚴重的衰落通道也能預測準確，且可以發現的是最佳因子的值和通道的時變速度、「信號多徑比」有很大的關聯。

Underwater acoustic communications differ from RF communications in two major aspects. One is the long multipath delay time covering tens to hundreds of symbols and the other is temporal variation of the acoustic channel at a time scale on the order of communication packet length. And the precision of channel estimation is the critical factor for the performance of channel estimation based equalizer. Here we using recursive least square algorithm as channel tracking algorithm. The RLS algorithm with a constant forgetting factor (FF) is not suitable for tracking time-varying channel because its convergence is slow when the FF is close to one, whereas the misadjustment is large when the FF is small. Therefore, the forgetting factor of RLS algorithm needs to be set adaptively in order to yield satisfactory performance in UWA environments. In this thesis, we provide a more directly method, say, from implementation of theoretical optimal value to set the FF, instead of controlling the forgetting factor based on the residual mean square error. The experiment result was presented based on ASIAEX data, and SMR was used as a measure for charactering the general quality of the channel. In the experiment result, we can observe that the proposed Forgetting Factor estimation is effective, even for severe fading channel. And it’s obvious that the value of optimal Forgetting Factor is highly correlated to the channel fading rate and SMR.