Fundamental frequency modeling method and system

Abstract

The invention discloses a fundamental frequency modeling method and a system. the method comprises steps: a rhythm layer is divided into a phrase layer, a word layer, a syllable layer, a phoneme layer and a state layer sequentially from high to low, wherein the phrase layer and the word layer are higher rhythm layers and the syllable layer, the phoneme layer and the state layer are lower rhythm layers; influences on fundamental frequency modeling of the higher rhythm layers by tone information in the syllable layer are determined; according to fundamental frequency features of the rhythm unit, an iterative method is adopted to build a fundamental frequency model from high to low layer by layer, and as for the higher rhythm layers, in the case of building of the fundamental frequency model, the influences on fundamental frequency modeling of the higher rhythm layers by the tone information in the syllable layer are removed. Thus, the influences on modeling of the higher rhythm layers by the tone information can be effectively eliminated, and the fundamental frequency features can be predicted more naturally.

Description

technical field [0001] The invention relates to the technical field of speech signal processing, in particular to a fundamental frequency modeling method and system. Background technique [0002] As one of the important features of the speech synthesis technology, the fundamental frequency feature includes not only the prosodic information of the short-term speech segment, but also the prosodic information of the long-term speech segment (prosodic information of the super-segment), such as tone information. How to predict the fundamental frequency feature more naturally is one of the important goals of speech synthesis effect. [0003] At present, the fundamental frequency modeling method generally adopted is the hierarchical fundamental frequency modeling method, which starts from the generation mechanism of the prosody and the additivity of the fundamental frequency characteristics in the log domain, such as formula (1) and figure 1 Shown: [0004] F0 all =F0 state +F0...

AI Technical Summary

Problems solved by technology

[0007] Embodiments of the present invention provide a fundamental frequency modeling method and system to solve the problem that the existing fundamental frequency modeling method cannot predict the fundamental frequency characteristics more naturally

Images