Supervisor: Jonas Gomes
Co-Supervisor: Gelson Mendonca (UFRJ)
For now the morph is done purely on the time domain. The results are already interesting, if not what expected. The next step is to use time-frequencies representations (like wavelet transforms) to acomplish the desired effect.
Like the image counterpart it is done by a combination of a "warp" ans a "fade". The placing of the "marks" affects directly the result of the morph. For now it is up to the user to place them correctly.
The result of a normal fade is at fade.au (12k), and as expected it is clear the presence of both sounds.
The morph at 50% is found at morph50.au (11k). The result, if not exactly what we would percieve as an intermediate voice, is one sound only and it's not possible to distinguish remains of either original.