| Résumé |
Voicing techniques are used by organ builders in order to make a pipe speak the way it should. The voicing art consists in mastering a sequence of steps. At each step, the organ-maker will change a particular parameter of the geometry of the pipe often by a slight amount in order to modify its actual sound. This process is mainly based on error and trial methods and empirical knowledge. In other words, the trajectory created by these steps can be sub-optimized. We have been interested to address this problem in the following way : is it possible to find an optimal trajectory (in the physical parameters map) supposing that we have a verbal description of the desired sound? This vast question will not find a definitive answer here. We will however try to show how useful can usual digital audio techniques such as time-frequency analysis and harmonics+residual decompositions be in that domain. This is assessed by results of listening tests administrated by way of computers. For example, it was shown that the voicer is naturally performing a kind of noise+harmonics separation and that a digital application of such a principle act for him as an audio-magnifier. Considering the starting transients (a difficult part of voicing), we will also show that it is possible to relate a verbal description (such as the onomatopoeia "hiss") to a particular set of tonal bursts revealed by time-frequency analysis. |
