Speech synthesis method and device, equipment and computer readable storage medium

Abstract

The invention provides a speech synthesis method and device, equipment and a computer readable storage medium, and relates to speech synthesis. The speech synthesis method comprises the steps: a reference speech sequence is determined, and a speech synthesis model and the target text vector corresponding to a to-be-synthesized target text sequence are obtained; the reference speech sequence is encoded through a reference encoder, and a target reference embedding vector corresponding to the reference speech sequence is obtained; the target reference embedding vector is subjected to style marking through a style mark layer, and a target style embedding vector corresponding to the reference speech sequence is obtained; and through a speech synthesis layer, based on the target text vector and the target style embedding vector, speech synthesis operation is executed, and target speech is obtained. The speech is synthesized jointly through the target style embedding vector and the target text vector, the speech rhythm synthetic speech expressed according to the target style embedding vector can be obtained, and the expressive power accuracy of the synthetic speech can be effectively improved.

Description

technical field [0001] The present application relates to the technical field of speech synthesis, and in particular to a speech synthesis method, device, equipment and computer-readable storage medium. Background technique [0002] With the rapid development of the TTS (Text To Speech, from text to speech) system, there are more and more scenarios using the TTS system, such as audio book reading, news reading, and chat assistants. Among them, the neural network model has the ability to synthesize Expressiveness affects the performance of speech, but in order to synthesize more human-like speech, neural network models must learn prosody, which is the combination of a set of phenomena in speech, such as paralinguistic information, intonation, accent, and style. [0003] Style contains a wealth of information, such as intent, emotion, and the intonation and voice flow that affect the speaker, so it is difficult to accurately define the style of speech, and the current TTS syst...

AI Technical Summary

Problems solved by technology

[0003] Style contains a wealth of information, such as intent, emotion, and the intonation and voice flow that affect the speaker, so it is difficult to accurately define the style of speech, and the current TTS system and end-to-end speech synthesis system can only learn one input The average prosodic distribution of the data, without style control, cannot synthesize speech with precise expressive power for longer text sentences

Images