- Had a meeting with the school social worker about what social skills we want the guy to work on this year and about how the groups will be set up.
- Got groceries.
- Got a flu shot.
- Checked out my last hunch about what's wrong with the vocode tool, but I couldn't really figure out a way to prove anything. I tried getting the real and imaginary results from the FFT by casting its result to a complex array, then putting it back together for the inverse FFT by putting complex numbers together from the real and imaginary results. The result (when applied to a simple decompose-recompose case) were fine, but that doesn't mean I'm correctly interpreting real and imaginary results because as long as you put them back together in exactly the opposite of the way you took it apart, it's always going to be fine. So, that was a waste of time. Had I looked at the JUCE code
first, I would have seen that the float array that you pass to it gets reinterpret_casted to a complex array, so it is actually correct to interpret the result that way.
- Chatted with a colleague about ways to survive in programming outside of corporate situations, the risk of saving everything for retirement, dealing with Jonathan Blow, and successful people losing their minds. At one point, someone with a username like [something]Dong joined the call, and we thought we were going to get zoom bombed with something horrible. Turned out that it was just someone showing up for the upcoming meeting, so we moved to another Zoom room.
- Made the vocode tool log signals as a bunch of numbers to text files.
- Made dinner.
- Went to MCG, heard some good morin khuur tracks, among other things.
- Had trouble comparing sets of 200,000 numbers to each other (though diffing helped with finding out when they were identical), so I tried graphing them in OpenOffice, but I couldn't trust it or anything I found online. So, I made my own tool for graphing text files full on numbers. Disturbingly, it may be revealing that I'm not even copying the source audio correctly? There's a lot of horizontal lines in the graphs (and lots of identical consecutive values in the text logs). Maybe I'll find the truth of the matter tomorrow.