The nature and perception of fluctuations in human musical rhythms

hands Date: 26.10.2011
Have you ever wondered why music generated by computers and rhythm machines sometimes sounds unnatural? One reason for this is the absence of small inaccuracies that are part of every human activity. Professional audio software therefore offers a so-called humanizing technique, by which the regularity of musical rhythms can be randomized to some extent.

But what exactly is the nature of the inaccuracy in human musical rhythms? Studying this question for the first time, we found that the temporal rhythmic fluctuations exhibit scale-free long-range correlations, i.e., a small rhythmic fluctuation at some point in time does not only influence fluctuations shortly thereafter, but even after tens of seconds. While this characterization is relevant for neurophysiological mechanisms of timing, it also leads to a novel concept for humanizing musical sequences. Comparing with conventionally humanized versions listeners showed a high preference for long-range correlated humanized music over uncorrelated humanized music. (Photo: Courtesy of Agbenyega Attiogbe-Redlich, www.hippocritz-school.com)

Press release of the Max Planck Society

Press release

Online survey

Here you can experience how humanization sounds when applied to the music of J.S. Bach. You can evaluate the quality of different humanizing techniques and thereby help us gathering statistical data (anonymously of course). Completing the survey takes approx. 10 min.

Participate in survey

Statistical results of the survey can be found here: View preliminary survey results

Examples of humanized audio samples

In order to listen to these audio samples, humanized with the "regular" (white noise) based technique and with our 1/f noise approach, please participate in the online survey. You will be able to listen to humanized versions (1/f humanizing and conventional white noise humanizing) of the following two musical excerpts:

J.S. Bach, Invention no.1 in C major, BWV 772

Bach Invention1

J.S. Bach, Prelude no. 2 in C minor (Well-tempered Clavier Vol. II, BWV 871)

bach Prelude 2

In this piece the first part shall be repeated once, note that due to the fluctuations added while humanizing, each line is unique, i.e. the repetition is not an exact repetition of the first part in case of the humanized pieces.

Pop genre example

A third audio example from the pop genre also shows the different humanizing techniques (the difference is audible mainly in the rhythm)

First, sample A, then sample B is played, separated by a 5 sec. pause. The two samples differ only in the rhythmic structure, all other properties such as pitch and timbre are identical.The first part (sample A) of this audio example was humanized by introducing LRC (using Gaussian 1/f-noise), while for sample B the conventional humanizing technique using Gaussian white noise was applied. The standard deviation σ_1/f = σ_WN = 15 ms of the time series of deviations used for humanizing was chosen rather large for the purpose of demonstration. The song was created and humanized in collaboration with Cubeaudio recording studio (Göttingen, Germany). We observed a clear preference for the 1/f humanizing over the white noise humanized version in our study.

Audio samples of humanized music can be found in the audio gallery.

----------------
Original publication:
H. Hennig, R. Fleischmann, A. Fredebohm, Y. Hagmayer, J. Nagler, A. Witt, F.J. Theis, T. Geisel, PLoS ONE 6(10): e26457 (2011). Download the article (Feel free to post comments and questions there).

~~~