正交分頻多工系統下運用壓縮取樣技術執行通道估測

Abstract

在正交分頻多工(OFDM)系統中，通道估測(channel estimation)經常是藉由安插在OFDM 符元(symbol)間的領航訊號(pilot)來完成。但是領航訊號的使用卻會影響到系統的效能，越多的領航訊號被安插在OFDM 符號間則能傳送的資料量就越少，系統的傳送速率(transmission rate)便會下降；此外，在某些系統中領航訊號的數量是有所限制的，因此如何利用少量的領航訊號來達到準確的通道估測便成了一個值得探討的問題。近年來，有研究提出一項名為Compressive Sampling (CS)的新技術，宣稱只需要運用少許的取樣值便能還原原始的訊號，只要該訊號本身擁有稀疏(sparse)的特性即可。而在時域上的通道響應其非零的位置通常不多，符合CS 技術的要求，因此我們可以將此技術應用在通道估測的問題上。在本篇論文中，我們提出使用一個Subspace Pursuit (SP)方法，證明其在通道估測方面比現存應用CS 技術的眾多方法有更佳的效能表現，並透過回授(feedback)的機制使此方法能在領航訊號密度很低的時候仍保有好的準確度，最後我們延伸此估測法至時變通道。透過模擬結果，可以看出我們所提出的通道估測法在高速移動的環境中依然擁有很好的效能。

In pilot-assisted OFDM systems, the channel estimation problem is usually solved by the using the pilot subcarriers inserted in OFDM symbols. However, more pilots used will lead to lower transmission rate, and the number of pilots is sometimes limited due to the systems. So we are facing a problem to accurately estimate the channel response while using a small number of pilots. Recently, a novel technique called compressive sampling (CS) has emerged, asserting to recover the sparse signals with a few measurements. Since the number of non-zero taps in time-domain channel response is small, we can then apply the CS methods to the channel estimation problem in OFDM systems. In this thesis, we propose using a subspace pursuit (SP) algorithm which is shown to be superior to the existing CS methods in channel estimation. The performance of proposed method is also shown to be good when pilot density is very low by adding a decision-feedback mechanism. Then, our problem is extended to the time-variant case. And simulation results show the proposed method performs well even when the speed of mobility is high.