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

Abstract

The embodiment of the invention discloses a speech synthesis method and device, computer equipment, a storage medium and a product, and the method comprises the steps: obtaining a text of a to-be-synthesized speech, and determining the type of the to-be-synthesized speech; performing fusion processing on the reference audio feature information corresponding to the voice type and a text unit in the text to obtain text voice feature information; determining a target duration prediction network according to the voice type; predicting audio duration information corresponding to the text unit according to the target duration prediction network and the text voice feature information; performing duration matching processing on the text voice feature information according to the audio duration information to obtain matched text voice feature information; and performing speech synthesis processing according to the matched text speech feature information to obtain target speech. According to the scheme, accurate text speech feature information can be extracted, and the corresponding duration prediction network is adopted according to the speech type, so that the synthesized target speech keeps the tone, rhythm and other information of the speech type, and the speech synthesis effect is improved.

Description

technical field [0001] The present application relates to the technical field of communication, and in particular to a speech synthesis method, device, computer equipment, storage medium and product. Background technique [0002] Speech synthesis technology converts text into corresponding audio content through certain rules or model algorithms, also known as Text to Speech (TTS), and its function is to convert text information generated by the computer itself or externally input Understandable, fluent voice and read it out. The traditional speech synthesis technology is mainly based on the waveform splicing method or the statistical parameter method. The splicing method needs to collect the waveforms corresponding to all pronunciation units in advance, and obtain the corresponding voice through waveform splicing. The statistical parameter method needs to model the spectrum characteristic parameters of the existing audio first. , to construct a mapping relationship between ...

AI Technical Summary

Problems solved by technology

[0003] Among them, the waveform splicing method needs to collect a large amount of audio for each speech type to cover all pronunciation units, and due to the splicing, the synthesized speech transition is unnatural, and the speech synthesis effect is poor. The statistical parameter method does not need to collect a large amount of audio audio, but due to the way of mapping, the synthesized voice has a strong mechanical sense and the synthesis effect is poor

Images