Voice Transformation

Diploma thesis (804 KB pdf)

The presented thesis deals with the transformation of the human voice. The focus is on the manipulation of general characteristics of speech signals (e.g. the sex or the age of a speaker).

The acoustic differences of speech signals of different groups of speakers such as men, women, young and old adults etc. have been determines and summarized on the basis of the available literature. With the help of two systems implemented in the software MATLAB, transformations of recordings of speech between the speaker groups and of voiced speech into whispering have been accomplished. The first system is based on spectral modeling synthesis, the second on is based on the resynthesis of single glottal pulses.

Furthermore, an overview over the literature about methods to map the voice of a speaker into the voice of a different person is given.