以最平半頻濾波器為基礎的一般性半頻濾波器設計

Abstract

在這篇論文，我們考慮以信號內差為應用的半頻濾波器的設計與特性。在以Bernstein多項式表示方法的基礎上，一個一般性半頻濾波器可以描述成以最平半頻濾波器的多相位元素的線性組合。所以內差可描述為片段近似的信號分解，我們會利用最小平方法來最佳化半頻內差濾波器的參數。因此，我們要預測的資料，可視為以最平濾波器組成的基底的線性組合。所以這些加權參數即可視為信號完整的表示。

The design and characteristics of the halfband filter s considered for signal interpolation. Based on the representation form derived from the Bernstein polynomia, a general halfband filter can be described as the summation of a set of polyphase comoponents of maxflat filters. Thus the interpolation is described as piece - wise signal decomposition, and we optimize the parameters of the halfband interpolation filter by minimizing the sum of the wquare interpolation error. Thus the data we want to predict can be viewed as a linear combination of the bases formed by the maxflat filters. So the weighting parameters can be viewed as a compact repressntation of the signals.