Research On HMM-Based Mongolian Speech Synthesis

With the development of society, information technology has made progress by leaps and bounds. Storage capacity and computing power of the new smart mobile devices have been greatly improved. In this context, traditional human-machine interaction with monotonous keyboard input and screen output can’t meet in people’s diverse needs. Therefore, some new human-machine interaction have emerged, of which the most prominent is the human-machine speech interaction, it mainly includes speech recognition and speech synthesis technology. Currently, human-machine speech interaction is so popular that has been widely applied in many smart mobile terminals of different fields. Meanwhile, researchers study the speech interaction is full of enthusiasm.As one key technology of human-machine speech interaction, speech synthesis experienced different development process. Today HMM-based speech synthesis method is the mainstream technology, Chinese, English, Japanese and other languages have developed the practical speech synthesis system based on this technology. However, Mongolian as a minority language, the research on Mongolian speech synthesis is still in its infancy. Thus, it is valuable to research HMM-based Mongolian speech synthesis technology.In this paper, HMM-based Mongolian speech synthesis technology was researched, the main work includes:design the rule of annotation Mongolian corpus data; identified the context attributes and corresponding questions associated with Mongolian for speech synthesis; studied the front text processing and prosody prediction of Mongolian speech synthesis system; achieve the HMM-based Mongolian speech synthesis system. Finally, we evaluated the synthesized Mongolian speech by subjective MOS method. Experimental results show that the synthesized Mongolian speech is fluent, naturalness and has high intelligibility. Thus proving the research work of this paper is valuable.