A Chinese syllable speech synthesis method and system based on improved Tacotron2 model

By improving the Tacotron2 model, a Chinese syllable speech synthesis model is constructed, and phoneme audio and syllable audio are directly processed, which solves the special pronunciation processing problem in Chinese choreography synthesis, and realizes efficient speech conversion and rich speech synthesis applications.

CN119049448BActive Publication Date: 2025-03-25HUAQIAO UNIVERSITY
2 Cites 0 Cited by

Patent Information

Application Number
CN202411555105.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-11-04
Publication Date
2025-03-25
Estimated Expiration
2044-11-04

AI Technical Summary

Technical Problem

The prior art is difficult to deal with the problem of special pronunciation in Chinese pronunciation synthesis, especially those syllables that do not correspond to Chinese characters.

Method used

Using the Chinese syllable speech synthesis method based on the improved Tacotron2 model, phoneme audio and syllable audio are directly processed by constructing the initial embedding module, encoding module, attention perception module, pre-network, decoding module, post-network layer and linear spectrum conversion layer to realize speech synthesis.

Benefits of technology

This method can effectively deal with Chinese syllable pronunciation without Chinese character carriers, improve the effect and efficiency of speech conversion, enrich the technology and application of Chinese pronunciation synthesis, and provide new tools and perspectives for linguistic research.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention discloses a Chinese syllable speech synthesis method and system based on an improved Tacotron2 model, and relates to the technical field of speech synthesis. The method comprises the following steps: constructing and training a Chinese syllable speech synthesis model; inputting the phoneme audio to be synthesized into the trained Chinese syllable speech synthesis model to realize speech synthesis; the Chinese syllable speech synthesis model comprises: a syllable audio initial embedding module, an encoding module, an attention perception module, a pre-network, an attention module, a decoding module, a post-network layer, and a linear spectrum conversion layer. The present invention can use the audio of initials and finals to synthesize the audio of the pronunciation of Chinese syllables without Chinese character carriers. Through this method, even if some pronunciations do not have corresponding Chinese characters, the corresponding syllables can be synthesized by combining initials, finals and tones, which can not only enrich the technology and application of Chinese speech synthesis, but also provide new tools and perspectives for linguistic research.
Need to check novelty before this filing date? Find Prior Art