盲頻道判別之階數決定及參數估測

Abstract

本篇論文探討二個盲頻道判別方法，第一個是子空間判別法[12]，第二個是最小平方近似判別法[11]。在最小平方近似判別法中，並未討論頻道階數未知的問題，也沒有其他論文針對這篇的頻道階數問題作探討。在這方面我們發現頻道階數與利用輸出訊號所形成的矩陣的秩有一規則的關係。所以我們先取一個比頻道階數大的值來當作階數值(超估測階數)，用這階數值來形成所謂的資料矩陣，並用已有的方法先估測出原本未加雜訊時的資料矩陣的秩。並且利用它來估測出頻道的階數值，然後估測出頻道的係數。最後，比較出最小平方近似判別法比子空間判別法的計算量要少。

In this thesis, we investigate two blind channel identification methods. The first is the subspace methods for blind channel identification [12], and the second is the least-squares approach to blind channel identification [11]. In the second method, it did not discuss the method of channel order determination. In this aspect, we find that there is a regular relationship between the channel order and the rank of the data matrix formed by output data signals. First, we take an order which is bigger than the true channel order (over-estimated channel order), and use it to form the data matrix. And we use an existent method to estimate the rank of the original data matrix. Then we use it to estimate the channel order and estimate the channel coefficients. Finally, we find that the computations of the least squares approach is less than that of the subspace methods.