Voice conversion using duration-embedded bi-HMMs for expressive speech synthesis | IEEE Journals & Magazine | IEEE Xplore