Today was mostly spent with Audacity, again. Started out finishing translating the FFT function from C++ to Java. Harder than it sounds. Also figured out exactly what each method was doing, and why. For some reason, bits have to be reversed, probably to do inverse function, and code had three ways of doing it, each faster than the last. After sorting out that out, figured out which of the two FFT shells was more useful. Since Axtell was working with simple version, focused on the logarithmic one, which was not any faster, but generated much clearer data. After all, human brains are configured logarithmically, not linearly. Didn't get it working yet.
However, did implement Hamming and Bartell windowing in Axtell's function. Made data somewhat clearer. Hopefully get Hanning to work soon, as, acording to Audacity, it provides the clearest graph and most obvious peaks.
Tomorrow: Write/find function to take data from FFT, find peaks, store them, and work on getting windows of a song to better analyze them.
Sounds great! I like the team work. I'm interested in the statement about human brains thinking logarithmically - point me to a source on that.
ReplyDeleteIf you look up harmonic series (music) in wikipedia you'll find more info about partials.