Modeling segmental duration in German text-to-speech synthesis | IEEE Conference Publication | IEEE Xplore