Corpus synthesis of Slovak speech

This work discusses a method of Slovak speech synthesis by concatenation of corpus speech units. The proposed synthesis uses diphones as elementary units. The paper is focused on unit selection mechanism that takes number of concatenations, word stress in Slovak and distance of chosen parameters of concatenated diphones into account. The proposed synthesis was implemented into MFC application using two methods of choosing candidate units for selection algorithms and two database formats. Comparisons of syntheses´ times of implemented methods and used database organizations are finally presented.