Re: autocorrelation

[Vinayak Dutt]

Ted Harding wrote:

#There's nothing wrong with the answer from fftconv. This computes
#a CONVOLUTION, which is defined as
#
#  Sum[over s] x(s) * x(t-s)
#
#i.e. one of the factors is taken in the reverse order.
#The CORRELATION [at lag t = -(N-1), .. 2, 1, 0, 1, 2, .. , (N-1)]
#which you want is defined by
#
#  Sum[over s] x(s) * x(s-t) = R(t) = R(-t)
#
#You can get it the right way roud by reversing one of the arguments
#to fftconv, if you wish. This gives the result you expected.
#

Yup. This is the correct explaination for the problem. I checked fftconv()
and found that it does padd to  length(Vector1) + length(Vector1) - 1.
So there is no need to pad as I had earlier suggested :-(



--vinayak-
/*
 * vinayak dutt
 * graduate student, ultrasound research
 * mayo graduate school, rochester mn
 *
 * e-mail: address@hidden
 *         address@hidden
 *         address@hidden
 *
 */
#include "disclaimer.h"