Consultation des notices

The existing algorithms for analysis, synthesis and transformation of sounds ask the users to have a high competence on signal processing. To make the access to these efficient algorithms easier, it is desirable to let the algorithms adapt themselves to the signal and to the transformation needed. One of the key parameter for nearly all the algorithms which analyze or process the sounds by using time-frequency representations or sinusoidal models is the size of the analysis window: here we show some examples about how this parameter is linked to the resolution of a signal analysis, and the information we get choosing a certain size instead of another. We are at present studying and developing algorithms for the automatic adaptation of the analysis window size. The first method approached is the parallel evaluation of multiple choices of the parameters and the subsequent combination of the final outcome through a comparison between their features.