If you want to see me write a really crappy FFT, read my FFT notes.

If you don't know what a Hann window is, read my windowing notes.

The problem with a *Fourier Transform* is that while we can find the
amplitudes of the different frequencies that are present, we can't know
at what exact time they occur. One way of visualising approximately how
amplitudes change through time is by taking lots of overlapping sections
of the signal we're analyzing, transforming those sections, and graphing
the results along a time axis. The result is called a *sonogram*
or a *voice-print*.

This is a plain sonogram of nF's guitar sample:

And here's one using the Hann window to taper sections of the signal before transforming them:

The Hann window reduces the spectral leakage a lot at the expense of lowering the resolution (the peaks become fatter). As always, you can grab the code. (sonogram.c)

[home]copyright © 2004 Emil Mikulic

$Date: 2004/10/09 08:47:34 $