Voice conversion method and system for non-parallel data

Abstract

The invention provides a voice conversion method and system for non-parallel data. The method comprises the steps of training a target speaker voice synthesis model through the large-scale synthesis voice library data except a source speaker and a target speaker; based on a text corresponding to the voice data of the source speaker, generating parallel data corresponding to the target speaker according to the target speaker voice synthesis model; training a frequency spectrum parameter conversion model by utilizing the voice data and the parallel data of the source speaker; and based on the voice data of the source speaker, generating the converted voice of the target speaker according to the frequency spectrum parameter conversion model and the target speaker voice synthesis model. According to the method provided by the invention, high-quality parallel data is forged by using the target speaker voice synthesis model, then a spectrum parameter conversion model is trained by using theparallel data, and voice conversion is carried out by using the spectrum parameter conversion model and the target speaker voice synthesis model, so that the conversion quality is ensured.

Description

technical field

[0001] The invention relates to the technical field of voice conversion, in particular to a voice conversion method and system for non-parallel data. Background technique

[0002] Speech conversion is a technique for modifying a source speaker's speech signal to match a target speaker's speech signal, so that it has the speech characteristics of the target speaker while keeping the speech information unchanged. The main task of speech conversion includes extracting and converting the characteristic parameters representing the personality of the speaker, and then reconstructing the converted parameters into speech. This process must not only ensure the clarity of the converted speech, but also ensure the similarity of the converted speech features.

[0003] In the existing speech conversion technology, most of the methods require two speakers to have parallel data (the text content corresponding to the speech is consistent). The main disadvantage of this meth...

Examples

Embodiment Construction

[0051] The preferred embodiments of the present invention will be described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described here are only used to illustrate and explain the present invention, and are not intended to limit the present invention.

[0052]The embodiment of the present invention provides a voice conversion method for non-parallel data, such as figure 1 As shown, the method performs the following steps:

[0053] Step 1: using large-scale synthetic sound database data other than the source speaker and the target speaker to train the speech synthesis model of the target speaker, wherein the large-scale synthetic sound database data includes text data and speech pair data;

[0054] Step 2: Based on the text corresponding to the speech data of the source speaker, generate parallel data corresponding to the target speaker according to the speech synthesis model of the target speaker, wherein the paral...

technical field

[0001] The invention relates to the technical field of voice conversion, in particular to a voice conversion method and system for non-parallel data. Background technique

[0002] Speech conversion is a technique for modifying a source speaker's speech signal to match a target speaker's speech signal, so that it has the speech characteristics of the target speaker while keeping the speech information unchanged. The main task of speech conversion includes extracting and converting the characteristic parameters representing the personality of the speaker, and then reconstructing the converted parameters into speech. This process must not only ensure the clarity of the converted speech, but also ensure the similarity of the converted speech features.

[0003] In the existing speech conversion technology, most of the methods require two speakers to have parallel data (the text content corresponding to the speech is consistent). The main disadvantage of this meth...

Examples

Embodiment Construction

[0051] The preferred embodiments of the present invention will be described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described here are only used to illustrate and explain the present invention, and are not intended to limit the present invention.

[0052]The embodiment of the present invention provides a voice conversion method for non-parallel data, such as figure 1 As shown, the method performs the following steps:

[0053] Step 1: using large-scale synthetic sound database data other than the source speaker and the target speaker to train the speech synthesis model of the target speaker, wherein the large-scale synthetic sound database data includes text data and speech pair data;

[0054] Step 2: Based on the text corresponding to the speech data of the source speaker, generate parallel data corresponding to the target speaker according to the speech synthesis model of the target speaker, wherein the paral...

Images