3 ms·

Can anyone translate this into plain English please? I have played the drums, and I still don't get it. I get that they are measuring the spacing and loudness o

Can anyone translate this into plain English please? I have played the drums, and I still don't get it. I get that they are measuring the spacing and loudness of his drumming, specifically how he hits the hi-hat instrument. As expected with any human player, there are fluctuations. He speeds up and slows down. He hits harder or softer. It sounds like they have found a pattern, and that's the part I don't get. These phrases seem central: 1. long-range correlated (LRC) fluctuations 2. detrended fluctuation analysis (DFA) Yet I can't figure out what these mean, even after reading about them in Wikipedia too.

anotheryou 9y ago
I can guess what 1. means: The hits are not exactly on time (fluctuating), but they are not randomly off either. If you observe a few hits you can guess something statistically about the next hits (correlated). These possible guesses span more than just 2-3 hits (long-range). I think they just observed timing. Not much novel insight, but they also made the hi-hat detection from a full-band recording automatic and hope this aids further research. The known insight is, that there are non-trivial, not yet understood patterns in the timing and it's proven that people prefer music that has these patterns. It seems that what feels right is neither something mechanically on point, nor something slightly off in a random fashion, nor just patterns within 2-3 neighboring hits, but something more sophisticated. What is really promising what they hint at at the end: Feeding these patterns in to drum computers. There are already "grove" patterns for an off timing, but they probably don't feel perfectly natural because they still lack these hidden patterns. --- About 2.: I'm really not sure, but I'd guess detrending would mean to calculate away things like an accelerando and only observe the pseudo random, minuscule changes in timing. An accelerando would allow for the easy guess that the next hits will come even faster, but this insight would be trivial. You want to exclude these trends from the calculations.

anotheryou 9y ago

I can guess what 1. means: The hits are not exactly on time (fluctuating), but they are not randomly off either. If you observe a few hits you can guess something statistically about the next hits (correlated). These possible guesses span more than just 2-3 hits (long-range). I think they just observed timing. Not much novel insight, but they also made the hi-hat detection from a full-band recording automatic and hope this aids further research. The known insight is, that there are non-trivial, not yet understood patterns in the timing and it's proven that people prefer music that has these patterns. It seems that what feels right is neither something mechanically on point, nor something slightly off in a random fashion, nor just patterns within 2-3 neighboring hits, but something more sophisticated. What is really promising what they hint at at the end: Feeding these patterns in to drum computers. There are already "grove" patterns for an off timing, but they probably don't feel perfectly natural because they still lack these hidden patterns. --- About 2.: I'm really not sure, but I'd guess detrending would mean to calculate away things like an accelerando and only observe the pseudo random, minuscule changes in timing. An accelerando would allow for the easy guess that the next hits will come even faster, but this insight would be trivial. You want to exclude these trends from the calculations.