The Statistic-Based Prosody Model of Mandarin Chinese

The prosodic component imply special information that cannot be expressed by letters, it is thus very important to model the prosody coinciding with context in Mandarin Chinese speech synthesizer. Some useful parameters are summed up to signify the context, and the focus of this paper is accent detecting. Since accents can not be found from text directly, tone-range ratio and duration of syllables are combined to detect them according to characters of Mandarin. Experiment has shown that the method introduced in this paper can decrease miss ratio and increase retrieve ratio.