Characterization of guitars through fractal correlation dimensions of initial transients
Rolf Bader · Journal of New Music Research · 2006
Three guitars are characterized using 96 initial transients each, recorded over four strings, playing with four notes on each string, apoyando and tirando and in three different loudnesses. These initial sounds have been analysed using the fractal correlation dimension algorithm, which shows the number of rules, that is the number of perceivable components, governing the sound. It is shown that the sound character of the instruments can be seen within the overall mean values and standard deviations of the different strings, attacks and volumes. The reasons for the different values are mostly inharmonic frequency components in the initial transient caused by resonant frequencies of the body. The minimum energy needed to drive these resonances could be a way for guitar-builders to determine the loudness and presence of their instruments. Also a pre-scratch-sound, the sound of a finger scratching along the string before releasing it, is shown to contain almost the whole overtone structure of the following sound and could be used to pre-determine the pitch or identify the instrument as a guitar.