Asked by Lisa Justin
on 1 Mar 2013

How can i estimate 75% overlap?

nfft=512

overlap=0.75

noverlap=nfft*overlap

is noverlap correct? I get error in my spectrogram

*No products are associated with this question.*

Answer by Wayne King
on 1 Mar 2013

Edited by Wayne King
on 1 Mar 2013

Accepted answer

You should specify the noverlap on the segments of your time data, not on the value of zero-padding you may or may not use for the DFT.

So base it on your WINDOW argument

WINDOW = 300; NOVERLAP = round(0.75*WINDOW);

Or if WINDOW is a vector,

NOVERLAP = round(0.75*length(WINDOW));

Also, you did not give the error you obtained, but I'm assuming that error is because 0.75 times an arbitrary integer is not necessarily an integer.

Still, what I said initially is what you should do. Base the number of overlapped samples on your window (segment) length, not on the value of your zero-padding. In many applications you may choose NOT to zero-pad the windowed Fourier transforms at all.

Opportunities for recent engineering grads.

## 0 Comments