A-005 Double Overlap Processing for Convolving a Long Impulse Response with an Infinite Length Signal

IPSJ

Bibliographic Information

Other Title
  • A-005 長時間インパルス応答と無限長信号のコンボリューションのための2重オーバーラップ法(A.モデル・アルゴリズム・プログラミング)

Search this article

Description

長いインパルス応答と無限に長い入力信号との畳み込みの手法を提案している.本手法は,従来インパルス応答が長い場合の畳み込み計算は,FFTの点数が大きくなるためにスーパーコンピュータを用いなければ計算が不可能になるといった状況であったのを解決する方法である.その問題を解決するため連続畳み込みを行う際に入力信号を小区間に分割するだけでなく,インパルス応答も小区間に分割し2重に重畳力日算法(overlap-add)や重畳保留法(overlap-save)を用いて畳み込みを行う方法である.考え方を述べて,数学的基礎を与え,使用例を示している.また,計算時間を通常のオーバーラップ処理と比較している.

Proposed here is a technique for convolving a long impulse response and an input signal of infinite length using the FFT algrithm of limited array sizes. Conventional overlap-add or overlap-save methods divide the input signal into small segments of the same length and integrates the convolved results of the impuls response and all the small segments of the input, by adding or saving the overlap portions at frame shifting. These methods, however, require a complex array of a length longer than or equal to the sum of the length of the impulse response and the length of one small segment of the input signal. So a large size array is required for convolving a long impulse response and a long input signal even if the input signal is divided into small segments, because of the length of the impulse response. If a user does not have facilities that satisfy the requirement, he/she has no other way than giving up the convolution. The algorithm proposed here solves the problem by dividing also the impulse response into small segments. So, the algorithm is named "Double Overlap" method. This report describes the idea and gives the mathematical basis of the algorithm, and then shows an example . Computational loads are investigated comparing that of the conventional overlap procedures.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top