Polyphone model training method, and speech synthesis method and device

Abstract

The invention discloses a polyphone model training method for speech synthesis, and a speech synthesis method and device. The method comprises the following steps of processing a voice data set and a text set so as to generate a training corpus set, wherein the text set corresponds to the voice data set, and the training corpus set comprises texts and Pinyin sequences corresponding to the texts; extracting feature information of the texts; and training polyphone models according to the feature information and the Pinyin sequence. According to the polyphone model training method for speech synthesis, in a polyphone model training process, manual labeling on Pinyin of the texts is not required, a training period of the polyphone models is greatly shortened, meanwhile, the circumstance that the trained polyphone models are inaccurate due to wrong manual labeling is avoided, and accuracy of the trained polyphone models is improved.

Description

technical field [0001] The invention relates to the field of speech technology, in particular to a polyphone model training method for speech synthesis, a speech synthesis method and a device. Background technique [0002] Speech synthesis, also known as text-to-speech (Text to Speech) technology, is a technology that can convert text information into speech and read it aloud. It involves multiple disciplines such as acoustics, linguistics, digital signal processing, and computer science. It is a cutting-edge technology in the field of Chinese information processing. The main problem to be solved is how to convert text information into audible sound information. [0003] In the speech synthesis system, the process of converting text information into sound information is as follows: first, the input text needs to be processed, including preprocessing, word segmentation, part-of-speech tagging, polyphone prediction, prosodic level prediction, etc., and then through the acousti...

AI Technical Summary

Problems solved by technology

[0009] For the first method, (1) it is labor-intensive, and the cycle of rule summary is long

Since there may be conflicts between the rules, after writing a rule, you need to spend energy to verify the effect to prevent conflicts between the rules, and then it is impossible to predict polyphonic characters

This process is extremely time-consuming, especially when there are many rules, it may take several months to manually summarize a series of rules that work well

(2) The rules have one-sidedness and limitations

The rules summarized in the early and later stages have certain contradictions and inconsistencies due to changes in the focus of consideration, which gradually lead to redundant rules, resulting in gradually lower quality, and it is difficult for humans to find problems and correct them

Or the rules only take into account the effects of a certain type of conditions, but in other contexts it will not apply

(3) The rules have problems of low scalability and low robustness

The problem of polyphone prediction depends on the results of front-end natural language processing modules such as word segmentation. Artificial rules can only be summarized and written based on the current word segmentation results. Once the front-end word segmentation results change in the future, the summarized rules may not be applicable.

[0010] For the second method, in the process of training the model, a large amount of manually labeled sample data is often required. Manual labeling of sample data takes a long time and is inefficient. Moreover, human errors may cause data quality to decline, thereby affecting all models. The effect of the polyphone prediction of the trained model

Images