H. Balgacem, M. Talbi, and A. Cherif (Tunisia)
Speech synthesis, pathology voice, Pitch, TD-PSOLA, DTW,
This paper describes the implementation of a speech synthesis toolbox using the TD-PSOLA method under MATLAB to facility the pathology voice transformation. This approach is based on the decomposition of the signal into overlapping frames synchronised with the pitch period. A modification of the pitch of a vocal signal with the help of an instrumental shifter will transpose the formants as well as the pitch, and thus alter the character of the voice. So to have sounds more comprehensible we can correct a pathological sound by moving the pitch, the formants, or both at the same time. To carry out the alignment of two files sounds, we must start by cutting out the source into a continuous list of diphones. We can make transformations on the second signal (the target) by concatenation of the diphones of the corpus. Then we align the source and the target by using the DTW technique. This procedure aims to correct the pathological voice as well as possible.
Important Links:
Go Back
