A method for combining intonation modelling and speech unit selection in corpus-based speech synthesis systems

In this paper, we focus on improving the quality of corpus-based synthesis systems by considering several candidate intonation contours. These candidates are generated by a unit selection procedure for which cost functions are defined. The consideration of several possible pitch contours adds an additional degree of freedom to the search for appropriate speech units. Objective and subjective tests confirm an improvement in the quality of the resulting synthetic speech.