Suprasegmental representations for the modeling of fundamental frequency in statistical parametric speech synthesis رسالة دكتوراه