在數位通訊系統及行動通訊系統中,通道之鑑定與等化對系統性能有 極重要的影響,而在近幾年來許多盲蔽式通道鑑定法已被提出,但大多數 之盲蔽式鑑定法是去鑑定一個由脈波訊號(pulse shape)與多路徑通道參 數(即衰減因子(attenuation factor)和時間延遲(path time delay))所 組合成的合成通道,其共同的最大困難之一是會受到一個共通之可鑑定性 問題的限制。 在本篇論文中,吾人將已知的脈波訊號資訊由合成通道中抽離出來, 並利用分數間距架構下的濾波器組(filter bank)表示法來對衰減因子及 時間延遲這兩種參數做一估計的工作,在此所使用的演算法為非線性最小 平方(Nonlinear Least Squares)參數估計法,吾人發現,利用此種演算 法可以使得一些先前不可鑑定的通道皆可被鑑定出來,且其性能亦優於傳 統之方法。
In many digital and mobile communication systems, channel identification and equalization play an important role in determining the system performance, and a lot of blind channel identification methods are proposed in recent years. But most of the blind channel identification methods are designed to identify a composite channel which is composed of pulse shaping filter and multipath channel parameters, i.e., attenuation factors and path delays. All of them are limited to a common channel identifiability problem when there exists common roots among subchannels. In this thesis, we extract the known pulse-shaping filter from a composite channel, and exploit the filter bank representation of the fractionally-spaced structure to estimate the parameters of attenuation factor and path delay. It can be accomplished by the nonlinear least squares estimated method. We find that some previous unidentifiable channel can be identified with our method. Furthermore, the performance is much better than some existing methods.