ASA/CAA '05 Meeting, Vancouver, BC

[ Lay Language Paper Index | Press Room ]

Identifying Musical Instruments Through their Microrhythms

Rolf Bader
Institute of Musicology, University of Hamburg,
Neue Rabenstr. 13, 20259 Hamburg, Germany

Popular version of paper 3aMU11
Presented Wendsday morning at 8 am, May 18, 2005
149th ASA Meeting, Vancouver, BC

If we listen to music, normally we do not have difficulty to recognize the instruments that are playing. The violin is different from the oboe; the guitar does not sound like a drum. We take this for granted, but what enables us to distinguish between these instruments? We propose an astonishing answer: the human ear may be able to hear the an instrument's distinct "microrhythm," its initial burst of sound (the "initial transient phase"), about 50 milliseconds (.05 seconds) from the beginning of the tone. How do we know? If we record an instrument sound and cut its initial transient, a saxophone may not be distinguishable from a piano, or a guitar from a flute. This is because the so-called "quasi-steady state" of the remaining sound is very similar among the instruments. But their tone beginning is very different, complicated and so very individual. But how do we perceive this initial phase?

The ear has three integration times in terms of musical sound. After about 5 milliseconds (ms), we are able to perceive more than just a click. This is related to the human inner ear capacity of building up the critical bandwidth for frequency discrimination. The second important time is about 50ms. Here, we begin to hear distinct frequencies, not very accurately, but our ear gives us the chance to perceive a pitch. Then after about 250ms, the whole sound is perceived very well. The pitch can be heard clearly and we have the possibility to discriminate fine structures within the sound. So how do we perceive the first 50ms?

The very beginning of a sound can just be perceived as a whole. We are able to analyze very brief sounds by using modern signal processing tools in the computer that break them down into their different frequencies. But still the ear is not really able to reach that level of precision. Here, we may ask, does the ear have other ways to distinguish musical instruments without making frequency calculations?

Taking 144 sounds of six different musical instruments, a guitar, a saxophone, a clarinet, a violin, a turkey saz and a balinese gender (metallophone), we have performed a microrhythmic analysis of the initial transients. Microrhythm here means the time period of neighboring amplitude peaks in the sound. These peaks could be understood as an impulse train reaching the ear. If this train would have a characteristic rhythm for each musical instrument, the ear could distinguish the sounds very fast without the help of the frequency components of the sound.

And indeed, different instruments have different occurences of microrhythmic relations with their standard deviation shown as vertical lines through the points. The figures below show three example plots of occurence of relations vs. relations. For example, if the time between one pair of amplitude peaks is twice as long as the time between the next pair of amplitude peaks, then the ratio is 1/2; if three times the length, then 1/3. The vertical axis shows how often such patterns occur.

The mean occurences of microrhythmic relations from 144 tones of a classical guitar are shown in this plot. It results in a beautiful figure showing preferred harmonic relations.


This is not true for the quasi-steady state of the guitar tone. Here more irregular relations occur.


Another example is the violin. Here we have more values on the upper half of the plot and the whole picture tends to more irregular relations.


If the ear would use these information - which we do not know yet if it does - it would be an effective way to recognize instruments by such short initial transients as about 50ms.

[ Lay Language Paper Index | Press Room ]