This Waterfall display is made up off successive FFT's overlayed on top of each other, with the most recent at the front and the previous data being offset to the back and to the right. The routines are fast enough to generate these images in real time as you listen to the sound source.

Displayed is the section of speech "and Access too."
In this example, the displayed data is restricted to that above -70dB, this helps us to identify the speech components by ignoring the less important frequency information.

These Waterfall displays are quite difficult to make sense out of, especially when you can't both hear and see the Waterfall display. They are very useful for examining and visualising speech signals.

Music signals in general are too complex to examine and interperate as a whole, but this tool provides a useful and interesting method of visualising sound (or any other sampled data).