The Method Development of Mandarin Vocal Imitation Base on Ai Neural Networks

In addition to user-friendly human-computer interaction, there is an easy-to-operate meaning at the operational level. With more and more applications of the interactive voice mode due to the advancement of artificial intelligence technology, pedestrianization is also intended to be applied The tone of the voice reply is more like a human voice. The most common voice interactive applications such as car navigation, mobile secretary (Android OK Google, iOS Siri, etc.), voice instant translation, most of the voice feedback tone are quite blunt, people feel that it is a robot, you can even say There is no emotion in it, and in the vocal part of Mandarin, the same sentence with different tones of cadence, indicating different meanings and more pronounced. Therefore, the project hopes to develop a system to design a Mandarin vocal imitation system for Mandarin voice. By integrating artificial intelligence techniques and collecting real voices, the relevant interactive voice applications can be more human-like feel.This project is expected to be implemented in three years. The first year will focus on how to convert speech into text / phonetic / pinyin and the establishment of vocal databases and Chinese corpus. The second year will use the first corpus and human vocals Database for text / phonetic / Pinyin to Mandarin vocal training module design; the third year of the program is combined with the results of the previous two years to complete the entire Mandarin vocal imitation system.

Project ID:PB10708-1641
External Project ID:MOST107-2221-E182-074