The grass is greener
I was walking back home at 10am a few days ago, happily exhausted, and I happened to notice a sloped lawn of some houses next to my residence. That grass was really green! I then started looking around at other grass -- the corner of a wooded area, stuff growing in cracks on the pavement -- and yep, it was all really green. Much greener than the grass in Vancouver!
Granted, the last time I was in Vancouver it was late summer; not precisely a time noted for having lots of rain. And come to think of it, I can remember looking out the kitchen window last year and being surprised at how green the lawn looked. (especially my parent's back-yard lawn)
But hey, it made a nice introduction to this blog post.
My teaching ended almost two weeks ago, and I immediately shifted into my "sleep whenever I want" schedule. This normally results in a 26-hour sleep cycle (awake 17 hours, sleep 9 hours), but for the first time ever, I ended up on a 14-hour cycle. Awake 9 hours, sleep 5!
I felt great, though. And it was useful to have the shorter cycle, because it meant that I spent some time at university by myself (so I could get work done), but also some time there when other people were around (so they could test what I was working on).
Oh, as for "what I was working on"... half the lab is now in either Australia or New Zealand, for a series of conferences and/or workshops. I was preparing an electronic clarinet that plays in 19-tone equal temperament and unearthing a two-year old version of rosegarden with a 19-tone equal temperament pitch tracker. I was happy to volunteer for the task, but I underestimated the amount of work required.
I'm now porting the ancient rosegarden pitchtracker code to the latest version (using Qt4 instead of Qt3). I'm happy to report that so far, it's going exactly on schedule: this job should take a week.
Granted, that's another week that I'm not doing research. Speaking of research, I'm seriously reconsidering whether to include a metronome in the rhythm game. About half the people I've showed the game to (and everybody over here) have complained about the metronome. Or rather, they've complained about having to follow the metronome. (yes, I'm sure there's a pianist or singer or whatever joke in there somewhere)
But there is a serious issue there. I did a few quick tests on myself -- I had the computer give me a 4-beat metronome introduction (set to a random number between 60 and 120 BPM), I tapped the rhythm, then the computer told me my tempo. I generally varied between -5 BPM and +10 BPM. Now, I'm a great ensemble player. I'm not claiming on being the best musician ever, but when it comes to rhythms, aligning rhythms, keeping track of what everybody's playing, etc., I'm above average for a classical musician with 20 years of training. So if I naturally vary the tempo by that much, there's no point penalizing students for doing the same!
Basically, "with metronome" tests your ability to play rhythms in a large ensemble (say, an orchestra). But when you're playing chamber music or in a private lesson, the tempo is much more flexible. I'd estimate that in a normal "simon says" type of private lesson (teacher claps a rhythm, then the student claps it back), the teacher wouldn't complain if the student was +- 20% of her initial tempo -- and probably wouldn't even notice if it was +- 10%. (yes, I should spend more time looking at psychology papers to find exact numbers for these)
So I'm now thinking that the most useful rhythm training tool would give an initial "target" tempo, but then stop the metronome and let people clap or tap the rhythm. Technically, this means that I need to detect the tempo of the student's exercise. This is a well-known problem in Music Information Retrieval, but this application has special constraints. First, it needs to be very reliable; we don't want to give students bad grades because the tempo-detection algorithm gives the wrong answer. Second, it needs to be robust against incorrect rhythms. If somebody taps four 8th notes instead of four 16th notes, the tempo detection should still work.
All in all, it's looking like the actual rhythm survey won't happen until January. This week is devoted to cleaning up the mess of previous research in the group (i.e. the rosegarden thing); next week I'll start the tempo detection in earnest.