Improved Smoothing

Just when I thought smoothing couldn't get better… I coded up three different smoothing algorithms for FuzzMeasure. Well, I actually used only two methods, but one method had two possible ways to implement it. Anyway, I wanted to use a fractional-octave based smoothing method, but it didn't work at first. I walked away from it at the time, and continued with a standard moving average filter. In testing, I noticed something really crappy. When the user chooses higher FFT sizes, the maximum smoothing amount does not work as "maximally" as with lower FFT sizes. This made perfect sense due to the increasing number of points. So, I needed to return to the fractional-octave based smoothing and figure out what I did wrong. It turns out that I misunderstood the algorithm slightly when I implemented it, and left a few bugs in the implementation to boot. Now, smoothing works wonderfully, and users have a much more useful metric which they can use to choose the smoothing amount. So, in FuzzMeasure 1.3, you can look forward to 1/48 octave up to 1/3 octave smoothing. Keep in mind — this does not mean the same thing as "1/3 octave analysis". I plan to add that later on.