Lightweight multi-speaker speech synthesis system and electronic equipment

Abstract

A lightweight multi-speaker speech synthesis system and electronic equipment, the system comprises: a text feature extraction and regularization module, a speaker feature extraction module, a feature fusion module and a speech generation module. The text feature extraction and regularization module is used to use a lightweight encoder to encode and extract the text information to be processed, and use a lightweight duration prediction network to perform each word corresponding to the text deep features output by the lightweight encoder. Or phoneme for duration prediction, and for length warping to obtain regular text features with the same length as the target mel spectrum. The speaker feature extraction module is used to generate features that can characterize the target speaker. The feature fusion module is used to fuse the features of the target speaker with regular text features. The speech generation module is used to perform deep feature extraction, dimension mapping, residual integration and speech generation on the fused features. The system supports multi-speaker speech synthesis and the synthesis speed is fast.

Description

technical field [0001] The present disclosure belongs to the technical field of speech synthesis, and relates to a lightweight multi-speaker speech synthesis system and electronic equipment. Background technique [0002] In recent years, neural network-based end-to-end speech synthesis systems have surpassed traditional statistical parametric speech synthesis systems in terms of system architecture and generated speech quality. End-to-end speech synthesis systems, such as the Tacotron2 system and the Transformer text-to-speech system (Transformer TTS system for short), directly use neural networks to convert text into corresponding speech, eliminating the need for a lot of complex text front-end processing, various Linguistic feature extraction, and complex domain expert knowledge. [0003] However, most of the current mainstream end-to-end speech synthesis systems use the attention mechanism to implicitly learn the text-to-speech alignment relationship, which brings a huge...

AI Technical Summary

Problems solved by technology

[0005] In addition, most of the existing text-to-speech systems can only realize a single speaker's single-style speech synthesis, and a few speech synthesis systems that can realize multi-speaker synthesis have the disadvantages of slow synthesis speed, large amount of calculation and memory consumption. question

Images