DocumentCode
1141120
Title
Systolic pipeline architectures for symmetric convolutions
Author
Bombardieri, John
Volume
40
Issue
5
fYear
1992
fDate
5/1/1992 12:00:00 AM
Firstpage
1253
Lastpage
1258
Abstract
Symmetries in the coefficients of common types of finite-impulse response (FIR) digital filters can be utilized to enable systolic pipeline realizations with considerable reduction in the number of multiplier stages. To obtain a concise description of the interconnections of these pipeline structures, a recursive state variable representation is presently developed. The representation is then utilized to examine the effects of symmetry on pipeline structures for discrete convolutions in one and two dimensions. For common types of one-dimensional filters, such as bandpass, phase-shift (Hilbert transform), and edge detector (differentiator) types, the number of multipliers may be approximately halved using odd/even symmetry. For common types of two-dimensional filters, such as bandpass, rho filters (tomographics), and edge detector (Laplacian) types, the number of multipliers can be reduced by approximately a factor of eight
Keywords
digital filters; pipeline processing; systolic arrays; FIR digital filters; Hilbert transform filters; Laplacian filters; bandpass filters; differentiator; discrete convolutions; edge detector; filter coefficients; finite-impulse response; multipliers; odd/even symmetry; one-dimensional filters; phase shift filters; recursive state variable representation; rho filters; symmetric convolutions; systolic pipeline architectures; tomographics; two-dimensional filters; Band pass filters; Convolution; Delay; Detectors; Digital filters; Equations; Finite impulse response filter; Phase detection; Pipelines; Tomography;
fLanguage
English
Journal_Title
Signal Processing, IEEE Transactions on
Publisher
ieee
ISSN
1053-587X
Type
jour
DOI
10.1109/78.134487
Filename
134487
Link To Document