Musical Transformations Using the Modification of Time-Frequency Images
Daniel Arfib, Nathalie Delprat · Computer Music Journal · 1993
The purpose of this article is to show how to make musical transformations with the help of time-frequency transforms of sounds. In order to make a parallel with the wavelet transform, we will use the name Gabor transform to specify a short-time Fourier transform on a rectangular grid. A Gabor timefrequency representation can provide one or more images from a sound and allows the reconstruction of this sound provided that certain conditions are fulfilled. The basic idea is to manipulate these images before their reconstruction. The resulting sound transform may or may not correspond to what was expected. To understand what is going on, we first look at the phase representation of very simple signals. Then we define what we call a legal image (in the sense that it is actually the transform of a signal) and explain how to turn an illegal image into a legal one. Finally, we demonstrate some deformations of one image, which leads to intimate sound transformations and some combinations of two images corresponding to a hybridization of two sounds.