Text information-based waveform concatenation voice synthesizing method

Abstract

The invention discloses a text information-based waveform concatenation voice synthesizing method. The method comprises the following steps: extracting acoustic parameters and text parameters of all primitives from original audio through sound segment segmentation, and training a time length prediction model and a weight prediction model according to the extracted parameters; performing layered pre-selection by adopting a layered pre-selecting method; performing primary pre-selection on the primitives in a language database by utilizing a target primitive analyzed from a text and the time length predicted by the time length prediction model to obtain candidate primitives; calculating the target primitive, the candidate primitives and the weight information predicted by the weight prediction model to obtain target cost; calculating the integrating degree between two adjacent primitives to obtain concatenation cost; searching the target cost and the concatenation cost by using a Viterbi search method to obtain the least cost path, so as to obtain an optimal primitive and a synthesized voice through smooth concatenation. According to the method, the voice synthesizing efficiency is improved; the real-time performance of the concatenation synthesis is enhanced; the prosodic features of the synthesized voice are improved.

Description

technical field [0001] The invention relates to a waveform splicing speech synthesis method, in particular to a text information-based waveform splicing speech synthesis method, which belongs to the field of intelligent information processing. Background technique [0002] Speech is one of the main means of human-computer interaction, and the main purpose of speech synthesis is to enable computers to generate high-definition, high-naturalness continuous speech. There are two main methods of speech synthesis. Early research mainly uses parametric speech synthesis, and the most commonly used synthesis method is based on Hidden Markov parameter speech synthesis method. As a specific implementation of statistical acoustic modeling method, this method performs Hidden Markov Modeling on the acoustic parameters of speech, reconstructs the trajectory of acoustic parameters through parameter generation algorithms, and finally invokes the speech synthesizer to generate speech waveform...

AI Technical Summary

Problems solved by technology

However, although this method can synthesize waveforms that are relatively close to the original speech, it is limited by the size of the corpus, and the stability of the synthesized speech is not ideal (the sound library is too large, the synthetic speech speed is slow, and it cannot be synthesized in real time; the sound library is too small. , synthetic speech is unstable), which greatly affects the sense of hearing

Moreover, the existing splicing synthesis system lacks consideration of the influence of text information on primitives when calculating the cost, and the synthesized speech is not very good in prosodic performance.

Images