Fast Convolution by Number Theoretic Transforms.

Abstract

The concepts involved in the efficient computation of discrete convolution by number theoretic techniques were reviewed using key portions of the applicable technical literature. First, discrete convolution and its computation by direct and fast Fourier transform (FFT) techniques are reviewed. The discrete Fourier transform (DFT), defined in a finite field or ring, and applicable number theory concepts, are used to describe the number theoretic transform (NTT) and its circular convolution properties. The considerations in the selection of a modulus that results in efficient modular arithmetic on a binary computing device are discussed. The limitations in transform length are presented, along with means for overcoming this restriction. The complex number theoretic transform (CNT) defined in both a finite field and ring is described. Finally, the computational efficiency of NTT vs FFT convolution is discussed.

Document Details

Document Type
Technical Report
Publication Date
Sep 12, 1975
Accession Number
ADA015276

Entities

People

  • Lawrence M. Leibowitz

Organizations

  • United States Naval Research Laboratory

Tags

DTIC Thesaurus Topics

  • Arithmetic
  • Complex Numbers
  • Computations
  • Computing Devices
  • Convolution
  • Discrete Fourier Transforms
  • Efficiency
  • Fast Fourier Transforms
  • Literature
  • Mathematics
  • Number Theory
  • Numbers

Fields of Study

  • Engineering

Readers

  • Parallel and Distributed Computing.
  • Structural Dynamics.
  • Theoretical Analysis.