Speech synthesis method, device and equipment and storage medium

Abstract

The embodiment of the invention provides a voice synthesis method, a device and equipment and a storage medium, which are used for improving the universality of voice synthesis and meeting the use requirements of common users without professional recording equipment and recording environments. The voice synthesis method comprises the steps that a voice broadcast instruction is received, and the voice broadcast instruction comprises a voice broadcast text and a target object corresponding to the tone used by the broadcast voice; a preset number of pieces of pre-collected voice data of the target object is acquired, voiceprint feature information of the target object is extracted by using a pre-trained voiceprint recognition model, the pre-trained voiceprint recognition model is generated bytraining based on the pre-collected voice data of the plurality of objects, and the preset number is smaller than a preset number threshold; to-be-played voice of which the tone is the tone of the target object is synthesized by using a pre-trained voice synthesis model based on a voice broadcast text, a pre-trained voiceprint recognition model and the voiceprint feature information of the targetobject; and the synthesized voice to be played is played.

Description

technical field [0001] The present invention relates to the field of voice interaction, in particular to a voice synthesis method, device, equipment and storage medium. Background technique [0002] The vehicle-mounted voice interaction system has always attracted public attention. A good voice-interaction system can not only improve the safety awareness of drivers and passengers, but also make the vehicle environment more intelligent. At this stage, star timbres are very popular among in-vehicle voice navigation interactions. This kind of timbre brings more entertainment effects, but customized timbres may improve the driver's safety awareness. Voice navigation is often accompanied by safety reminders, such as "fasten your seat belt" and "slow down ahead". If you customize the navigation voice to your parents, partner or your own children, the driver will be more willing to listen to it even if it doesn't matter. These "friendly" safety reminders will also enhance the sens...

AI Technical Summary

Problems solved by technology

[0004] At the same time, the multi-speaker technology based on end-to-end TTS has also been greatly developed. On the basis of the existing end-to-end TTS, scholars add audio tags of multiple speakers to distinguish them, and then conduct multi-speaker training. According to the numbers of multiple speakers, which voice can be used to synthesize the current text and realize the flexible switching between different speakers, it has certain practical value, but there is a big limitation, that is, the model needs a large number of speakers. Speaker data, each speaker needs at least several hours of professionally recorded, high-quality voice data to ensure the quality and practicability of the model, and it is not universal. For ordinary people, there is usually no professional recording equipment And the recording environment, and the training process usually requires on-site supervision by a special person, and the recording is repeated repeatedly to ensure the recording quality. Ordinary people do not have enough time to record such high-quality training audio for a long time, which makes the model unable to be used by ordinary people. scope

[0005] To sum up, the speech synthesis methods in the prior art are not universal, and cannot meet the needs of ordinary people who do not have professional recording equipment and recording environments.

Images