The Synthetic System Of Text-driven Lip Parameter Of Tibetan Xiahe Dialect

This paper use matlab to structure the parameter extraction platform, platform extractes the10kinds of lip animation parameters to describe the lip from the lip avi video. The initials and finals of Xiahe are divided16kinds of basic static lip phoneme with the method of clustering.At last,this paper gets phoneme time information by marking,then achieves Xiahe lip synthetic.The main contents are as follows:First,according to the characteristics and pronunciation of Tibetan Xiahe words Voice Features to establish the Xiahe lip model and a platform for parameter extraction. The platform read avi video files, extract each frame of video image, and can be labeled in the image frame, the extraction of lip parameters, and to extract the parameters to save.Secondly,design algorithm to Cluster Xiahe Initials and finals by the parameters extracted from the platform parameter,we get15kinds of Xiahe lip phoneme and16kinds of the Tibetan Xiahe basic lip.Third, based on the basic lip and text, time information to achieve the Tibetan Xiahe parameter synthesis systems, system according the lip speech mapping table and the text-to-speech conversion table to synthetic Xiahe, then realize the Image and voice sync playback.Finally, in order to improve the naturalness of synthetic lip,we do the synthesized lip post-processing. And we use the Perception experiments and accumulated evaluation to test synthesis effect.In summary, this paper based on the characteristics of Tibetan speech, relize the lip parameter extraction system of Xiahe dialect and analysis the parameters to get the static lip phoname.This paper relize the synthetic system of text-driven lip parameter of the Tibetan Xiahe dialect and the synthetic effect have been tested. The test results show the effect of synthesis system error within the reasonable range, synthetic method is feasible and effective.