Speech synthesis method and device

Abstract

The invention discloses a voice synthesis method and a device. The voice synthesis method comprises steps of performing text characteristic extraction on a text to be synthesized to obtain the context characteristic information, obtaining a pre-generated model, wherein the pre-generated model is generated by training according to the context characteristic information of the training sample and converted acoustic parameter, and the converted acoustic parameters comprise a plurality of rhythm level fundamental frequency parameters, determining the model output parameter corresponding to the context characteristic information according to the model, wherein the model output parameters comprise a plurality of the rhythm level fundamental frequency parameters, performing the fundamental frequency reconstruction on the plurality of rhythm level fundamental frequency parameter, and synthesizing voice according to the parameter after the fundamental frequency reconstruction and the other parameters in the model output parameters. The method can improve the performance result of the synthesized speech.

Description

technical field [0001] The invention relates to the technical field of speech synthesis, in particular to a speech synthesis method and device. Background technique [0002] Now people are not only satisfied with the clarity and intelligibility of synthesized speech, but also require the synthesized speech to have better naturalness and expressiveness. In natural speech, fundamental frequency is the main factor affecting naturalness and expressiveness, so the accuracy of fundamental frequency modeling directly affects the naturalness and expressiveness of synthesized speech. [0003] In traditional speech synthesis systems, fundamental frequency modeling uses a multi-space probability distribution hidden Markov model (multi-space probability distribution HMM, MSD-HMM) modeling method, which can be very good for state-level, The fundamental frequency contour (or trend) of the consonant and final level is modeled, but it is difficult to learn the higher-level fundamental freq...

AI Technical Summary

Problems solved by technology

[0003] In traditional speech synthesis systems, fundamental frequency modeling uses a multi-space probability distribution hidden Markov model (multi-space probability distribution HMM, MSD-HMM) modeling method, which can be very good for state-level, The fundamental frequency contour (or trend) of the consonant and final level is modeled, but it is difficult to learn the higher-level fundamental frequency trend of words, phrases or sentences, which makes the rhythm and expressiveness of the synthesized speech insufficient

Images