Feature selection for the prosody modelling of synthetic speech | IEEE Conference Publication | IEEE Xplore