Vocoder based on residual signal spectrum reconfiguration

Abstract

The invention provides a vocoder based on residual signal spectrum reconfiguration. The vocoder based on residual signal spectrum reconfiguration comprises a parameter extracting module, an inverse filtering module, a periodic signal branch, a nonperiodic signal branch and a parameter voice synthesizer module. The parameter extracting module is used for receiving to-be-analyzed voice data and calculating fundamental frequency parameters and frequency spectrum parameters. The inverse filtering module is used for designing an inverse filter according to the frequency spectrum parameters to inversely filter the voice data to obtain residual signals. The periodic signal branch is connected with the inverse filtering module and used for generating periodic excitation signals according to the fundamental frequency parameters and the residual signals after inverse filtering. The nonperiodic signal branch is connected with the inverse filtering module and used for generating nonperiodic excitation signals according to the fundamental frequency parameters and the residual signals after inverse filtering. The parameter voice synthesizer module is connected with the periodic signal branch and the nonperiodic signal branch and used for calling a voice synthesizer to synthesize voices according to the periodic frequency spectrum parameters, the periodic excitation signals and the nonperiodic excitation signals. The residue signals after inverse filtering is modeled to enable the excitation signals to be higher in accuracy, and quality of synthesized voice is improved.

Description

technical field [0001] The invention relates to the technical field of signal processing in the electronic industry, in particular to a vocoder based on spectral reconstruction of a residual signal. Background technique [0002] Vocoder technology is an important part of harmonious human-computer interaction technology. It can process a person's voice so that it can synthesize natural and smooth speech. Its research results have important implications for speech coding technology, parametric speech synthesis technology, The development of human-computer interaction technology is of great significance. [0003] However, the current vocoder technology uses a pulse plus white noise excitation model, which uses a pulse train or Gaussian white noise to construct an excitation signal. Due to the strong periodicity of the pulse train, the sound quality of the synthesized speech using this excitation model is not high. [0004] In addition, the traditional periodic spectrum calcul...

AI Technical Summary

Problems solved by technology

Due to the strong periodicity of the pulse train, the sound quality of the synthesized speech using this excitation model is not high

[0004] In addition, the traditional periodic spectrum calculation method based on spectral parameters can only fit the formant structure of the spectrum, resulting in the loss of the fine structure of the spectrum

The periodic excitation modeling method based on the time domain often causes the loss of high-frequency band energy during periodic sampling

Periodic spectrum estimation modules based on codebooks tend to be over-smoothed when selecting code values

Images