Prediction of word prominence.

摘要: Control of prosody is essential for the synthesis natural sounding speech. Text-to-speech systems tend to accent too many words when taking into account only distinction between open-class and closed-class words. In prominence-based approach [1], degree accentuation a syllable described in terms gradual prominence parameter. This paper presents calculation level based on their word class, classes surrounding words, position clause. Rules predicting are derived from statistical analysis prosodic database. The hand-crafted rules compared with results several machine learning algorithms same material. Furthermore, perceptual test an resulting speech signals carried out.