有限長データ用ローパスフィルタの研究 : 一般化ローパスフィルタと点対称拡張法を用いた空間型・周波数型ローパスフィルタ

概要

論文の詳細を見る
A low-pass filter is applied mostly to finite length data. Because of this, unexpected vibration occurs at the end of data as an output of the low-pass filter. To solve this problem, use of a Gaussian regression (GR) filter and a point symmetric extension method has been proposed. However, the former method is difficult to handle because its transmission characteristic differs from that of the low-pass filter, and the latter method cannot be applied when the data end shows an abnormal value. Therefore, in this paper the following three methods are proposed as a solution to the vibration problem that occurs at the data end. The first method is to use a generalized low-pass filter of which regression curve's coefficient is fixed so that the same transmission characteristic as that of the low-pass filter is obtained. The second method is to use a spatial low-pass filter to which a point symmetric extension method is applied after replacing the data end with the output of the generalized low-pass filter so that the abnormal value at the data end is accommodated. The third method is to use a frequency low-pass filter that provides the same output as the second method does. All these three methods solve the data-end vibration problem, and provide the same output as that of a low-pass filter when data other than at the data end are applied.
社団法人精密工学会の論文
2005-12-05