Expressive speech synthesis for a Radio DJ using Vocaloid and HMM's

TitleExpressive speech synthesis for a Radio DJ using Vocaloid and HMM's
Publication TypeMaster Thesis
Year of Publication2013
AuthorsFloría, H.
AbstractThis master thesis deals with the modeling the speaking style of a Radio DJ speaker. In the context of the radio DJ speaker we designed a corpus that represents how radio DJs use to present songs being played on a radio show. A professional speaker has been recorded reading these radio shows into two different expressive styles. Style dependant models are trained with the recorded data. These can then be used to generate the expression parameters (syllable duration, pitch and dynamics) for any message to synthesize. The system is then modified by adding contextual information which is taken into account to train the models. The system is evaluatued to show whether this modification of the models helps to improve the speaking style models.
intranet