Speech synthesis for embedded system based on initial and final units

摘要

To make corpus meet the demand of speech synthesis in embedded systems, this paper introduces the initial and final units in a speech synthesis system for Chinese to enhance the compressibility of voice database. According to the characteristics of Chinese pronunciation, the initial and final units are classified and the time factor, fundamental frequency factor, intensity factor are brought in to improve the splicing models. Results of the frequency spectrum, intensity spectrum and mean opinion score (MOS) show that this method can meet the basic needs of speech synthesis in many embedded systems.