On Speech Synthetic Units Considering Prosodic Parameters