Singing synthesis method and device, computer device and storage medium

Abstract

The invention discloses a singing synthesis method and device, a computer device and a storage medium, and is applied to the technical field of computers. The method provided by the invention comprises the following steps of acquiring the music data of a target song, and encoding the music data through a preset encoder to generate the music encoded data; inputting the music encoded data into an attention mechanism network for coding, and generating the feature information; inputting the feature information into a singing synthesis model, and generating the first sound spectrum data; and decoding the first sound spectrum data to obtain the first linear spectrum data, and converting the first linear spectrum data into a target song audio in a Fourier transform mode. According to the present invention, the singing synthesis model is obtained through training a small amount of training data, the synthesis efficiency of the target song audio is improved, and meanwhile the synthesis effect of the target song audio is guaranteed.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a singing voice synthesis method, device, computer equipment and storage medium. Background technique [0002] With the development of computer technology, speech synthesis technology is becoming more mature, especially in transportation and education. It has been widely used in medical and social fields. Such as traffic voice navigation, language learning pronunciation, medical voice control, etc., which facilitate people's work and life. [0003] A further singing voice synthesis system (Singing Voice Synthesis, SVS) is a system that generates natural singing voices based on given score and lyrics information. SVS requires the duration and pitch of each syllable to be controllable. [0004] The current singing voice synthesis system adopts the method of deep neural network, and the audio data of the target singer is used as training data, which is input into the neural net...

AI Technical Summary

Problems solved by technology

[0005] However, this method requires a large amount of training data, and can only be trained for the same singer, and the synthetic audio of the same singer's voice can be obtained

This method leads to low applicability of the singing voice synthesis system, and requires a large amount of training data, and the efficiency of synthesizing singing voice is not high

Images