Synthesis of Human Voice Fragments Based on Frequency Spectra Reconstruction

Abstract

This article details the development of methods for the synthesis of phonemes of the human voice based on the analytical description of individual formants. A technique for analyzing the spectrum and spectrograms of original phonemes to obtain the main amplitude-frequency characteristics of the signal components is presented. An algorithm to reconstruct a speech signal based on the obtained sets of parameters is proposed. A technique to assess the quality of synthesized speech elements is described