View Full Version : answers to Cross-cor

Rosa M Angulo-kinzler
09-02-1994, 01:17 AM
Here are the answers I received to my question about cross-correlation
and cross-FFT software.

************************************************** ********
From: Joergen Assentoft
Subject: Re: cross-FFT or cross-correlation

Try the company: MOMENTUM Data System, Phone 7145576884, FAX: 7145576969.
The software can run on WINDOWS and SUN Sparc. You have also the
possibility the make real-time A/D or D/A conversion and use a PC or SUN
Sparc Board with a DSP-processor.

************************************************** ********
From: mgm@nwu.edu (Mitchell Gil Maltenfort)
Subject: Re: cross-FFT or cross-correlation

Are you looking for packages or just useful bits of code? I have an FFT &
cross-correlation routine I wrote that's fairly clear and reliable.

Also, you might want to bear in mind that you can estimate the
cross-correlations of velocity and/or acceleration from the
cross-correlations of position. Since differentiation (and integration)
are linear operators, you can play some interesting games. Check Bendat
and Piersol's book "Random Data" and similar references.

Personally, I don't trust coherence. It's basically the correlation
coefficient of components at different frequencies, and assumes that X and
Y are related by a linear system. Cross-correlation may give you a clearer
result - besides the points I raised above.

Mitchell Gil Maltenfort Northwestern University Chicago, Illinois

************************************************** ********
From: richard@HMS01.HMS.UQ.OZ.AU
Subject: Re: cross-FFT or cross-correlation

With respect to the subject of your BIOMCH-L posting, I have previously used
the statistical package BMDP to perform bivariate spectral analyses
(cross-FFT). It wasn't something I pursued in great depth at the time,
although it did seem to offer a lot of potential. I have thought of applying
the technique to some of the data sets we are currently analysing.
I have used both the Sun (SPARC) and PC versions of the software. I
understand that academic prices are available. Simon Fraser University had a
site licence for the Unix version of the program, which we able to install
on our own machines in the lab. It is a very diverse set of routines, but
they are generally quite straightforward to use. Although you could also
achieve the same result with a package like MATLAB, this would involve a
good deal more development. My impression is that BMDP would provide the
information you require.
Richard Carson.

************************************************** ********
From: Graeme Wood
Subject: Re: cross-FFT or cross-correlation

I've used a commercially available package named TSA (Time Series Analysis)
to perform cross-spectral analyses of EMG signals. Your application would be
no different in principle. The package costs around AUD 600, and is
available from Dr John Henstridge (the author) at Unit 8, 154 Hampden
Road (P.O. Box 630), Nedlands, WA 6009, Australia. (Fax: +61-9-386 3202).

Graeme A. Wood

************************************************** ********
From: YUB@rcf
Subject: RE: cross-FFT or cross-correlation

Hi, Rosa,
I think calculating the coefficient of multiple correlation between the right
and left leg, and between your normal and abnormal subjects may be one of the
techniques you can use. The details of the calculation of the coefficient of
multiple correlation can be found in "Repeatability of kinematic, kinetic, and
electromyographic data in normal adult gait" by Kadaba and et. al. in Journal
of Orthopaedic Research, Vol 7, No. 6, 1989. The calculation of this
CMC is very simiple. I believe you can write a program to do it without any
difficulty, or you can use a statistic software package to do it.
Good luck.

Bing Yu, Ph.D.\

************************************************** ********
From: Gary Kamen
Subject: cross-correlation software

If you stay in the time domain, you should be able to do cross-correlations
with one of the statistical packages, like BMDP or SPSS. I have used
a signal processing package called HYPERSIGNAL (Hyperception, Inc.,
9550 Skillman LB125, Dallas, TX 75243, tel:214-343-8525) and this should
also work, as well as handling the problem in the frequency domain.

Gary Kamen
Boston University

************************************************** ********
From: Jill Whitall
Subject: Re: cross-FFT or cross-correlation

We (JoAnne Lazarus and I) have done cross-correlations using the
Systat commercial package...very easy to use...let me know if you need
more information..

************************************************** ********
Thanks to all of you!
Rosa M. Angulo-Kinzler