On the transformation of the speech spectrum for voice conversion

In many speech applications, control of the speech individuality is required. These applications include the personalization of the voice of speech synthesizers, the restoral of voice individuality for interpreting telephony, the improvment of abnormal speech intelligibility. It is generally admitted that both prosodic and spectral parameters have to be changed in order to modify the speech individuality. Several algorithms have recently been proposed for the spectrum control. This paper presents some improvments added to these previously proposed methods and compares 4 approaches in the same common framework of voice conversion for application to text to speech synthesizers.