Histograms of synthesised attributes can be calculated as well as the
raw data attributes. Two obvious ones will be to do the histograms on
the 
and 
features calculated above. This is
basically a ``poor man's''
energy spectrum, since it gives an estimate of the distribution of the
acceleration and velocity of the sign.
To understand how this gives us an energy spectrum, consider a sign
with smooth, slow, continuous motion. The distance covered will be
uniform, so there will be a ``peak'' in the 
with a low
reading, and have a very low 
reading, since acceleration
is constant. A smooth, fast sign will have a peak in the 
at
a higher reading level, but a similar 
histogram. Consider
a movement that has shaking or waving in it; this will have a
distinctive 
histogram, with peaks that will indicates
periods of high acceleration. The 
would show very little
movement in its histogram, because it is about the same point.