Multi-to-multi speaker conversion method based on STARGAN and x vector

Abstract

The invention discloses a multi-to-multi speaker conversion method based on STARGAN and an x vector, which comprises a training stage and a conversion stage, wherein a speech conversion system is achieved by combining the STARGAN and the x vector, the personality similarity and quality of the converted speech can be greatly improved, particularly, for the short-time utterance, the x vector has better characterization performance and better speech conversion quality can be achieved, meanwhile, the problem of over-smoothing in C-VAE can be overcome, and a high-quality speech conversion method isachieved. In addition, the method can achieve the speech conversion under the condition of non-parallel text, the training process does not need any alignment process, the universality and practicability of a speech conversion system are improved, and the method can also achieve that the conversion system with multiple source-target speaker pairs is integrated in a conversion model, namely, the multi-speaker-to-multi-speaker conversion is achieved, and the system has a better application prospect in the fields of cross-language speech conversion, film dubbing, speech translation and the like.

Description

technical field [0001] The invention relates to a many-to-many speaker conversion method, in particular to a many-to-many speaker conversion method based on STARGAN and x-vector. Background technique [0002] Speech conversion is a research branch in the field of speech signal processing, which is developed and extended on the basis of speech analysis, recognition and synthesis. The goal of voice conversion is to change the voice personality of the source speaker so that it has the voice personality of the target speaker, that is, to make the voice spoken by one person sound like another person's voice after conversion, while preserving semantics . [0003] After years of research on voice conversion technology, many classic conversion methods have emerged. These include most voice conversion methods such as Gaussian Mixed Model (GMM), Recurrent Neural Network (RNN), and Deep Neural Networks (DNN). However, most of these speech conversion methods require that the corpus u...

AI Technical Summary

Problems solved by technology

However, the identity label of the speaker in this method cannot fully express the individual characteristics of the speaker, so the converted speech has not been greatly improved in speech similarity

Images