New algorithms for convolutions and FFTs.
Item
-
Title
-
New algorithms for convolutions and FFTs.
-
Identifier
-
AAI9207098
-
identifier
-
9207098
-
Creator
-
Liu, Hongyi.
-
Contributor
-
Adviser: Richard Tolimieri
-
Date
-
1991
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Computer Science | Engineering, Electronics and Electrical
-
Abstract
-
The FFT algorithm with prime size is a key to develop the general FFT algorithms. The basic tool of computing the prime size FFT is the use of the cyclic convolution algorithm. Unfortunately, in many cases, the traditional algorithms for cyclic convolution are not very efficient and the computational structures with these algorithms are not flexible for the parallel implementation on different architectures.;By introducing the new concept half-cyclic convolution, the cyclic convolution can be treated as a special case of the new concept. Some half-cyclic convolution algorithms with general case and some special cases have been designed. The new algorithms for cyclic convolution with different purposes then can be developed, which can be very efficient and well-structured for both sequential and parallel processing. In particular, the prime case FFT algorithms can be improved a lot by using these new cyclic convolution algorithms. In addition, the algorithms for multi-dimensional FFT and some special computations such as symmetrized FFT can also be improved or redesigned. Furthermore, most of the algorithms for convolution and FFT could benefit from this development.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
