Voice conversion method under asymmetric corpus condition on basis of adaptive algorithm

Abstract

The invention discloses a voice conversion method under the asymmetric corpus condition on the basis of an adaptive algorithm. According to the method, firstly, source speaker and target speaker models are respectively obtained through training from a reference speaker model via utilizing a few of training sentences by a MAP (maximum a posteriori) algorithm; then, Gaussian normalization and average conversion methods are respectively provided through utilizing parameters in an adaptive speaker model; in addition, in order to further improve the conversion effect, a method combining the Gaussian normalization method and the average conversion method is further provided; and meanwhile, because the training sentences are limited, the accuracy of the adaptive model is inevitably influenced, the invention provides a KL (Kullback-Leibler) divergence method, so that a speaker model is optimized during the conversion, and subjective and objective experimental results show that the frequency spectrum distortion degree, the converted voice quality and the target voice similarity are respectively improved. All the methods provided by the invention obtain the effect similar to that of a classical GMM (Gaussian mixture model) method under the condition on the basis of a symmetric corpus.

Description

technical field [0001] The invention relates to a voice conversion method based on an adaptive algorithm under the condition of an asymmetric corpus, and belongs to the technical field of voice signal processing. Background technique [0002] Speech conversion refers to a technology that converts one person's speech characteristics into another person's speech characteristics, while keeping the semantic content unchanged. It has a very wide range of applications: such as speech synthesis for personalization; speech communication at low bit rates; restoration of speech damaged in medicine, etc. Speech-changing technology has come a long way in the past few decades. A series of speech conversion methods represented by codebook mapping, Gaussian mixture model, neural network and other methods have appeared. These methods achieve the conversion of the speaker's voice personality characteristics to a large extent. However, these methods mainly focus on speech conversion based ...

AI Technical Summary

Problems solved by technology

However, there are many defects in these methods: for example, the maximum likelihood bilinear regression method relies on the pre-prepared conversion function obtained from the training of the symmetric corpus; the bilinear transformation method requires a large number of training sentences of the source speaker and the target speaker to ensure The accuracy of conversion; the nearest neighbor loop iteration method is based on the fact that the nearest adjacent spectral features correspond to the same phoneme, and at the same time requires a large number of training sentences

Therefore, the above methods are difficult to implement in practical applications and are not easy to operate

Images