Re: A tiny problem ...

[Kire Pudsje]

On Wed, May 3, 2017 at 7:58 PM, Dr.-Ing. Dieter Jurzitza <address@hidden> wrote:

Hello Kire,
hello Doug,
thank you for the feedback and for taking the time to reply.

It does not look like I have overlooked a vectorizing option in the code I
wrote, though - according to your input Kire - there might be a way to do this
in the frequency domain what could be faster because the number of iterations
could be reduced.

Honestly speaking I did not understand what Doug was suggesting ...

But anyway, the code runs - though a little slower than it could probably be
running, and the results I get are the way I expected them to be.
Regards

Have a look at https://en.wikipedia.org/wiki/Autocorrelation#Efficient_computation

which shows it for autocorrelation. From there it is a small step to covariance.

There are two tricks involved:

A convolution in time domain is just a multiplication in the frequency domain.

A time shift will become an exponential in the frequency domain.