Análise e síntese da fala por interpolação de ondas

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

This doctoral dissertation addresses to study of waveform interpolation (WI) for speech coding, and the description, visualization and implementation of an algorithm to simulate a conventional WI system of analysis synthesis of speech signal, like portion of a WI system for coding - decoding that operates on the outer layer which is called of WI analysis synthesis system (standard). From this system researches are performed looking for improvement on the localization and extraction process of characteristic waveforms (CWs), that are obtained through interpolation at regular positions at the analysis stage. The propositions for improvement are presented, where part of them were implemented resulting on the system called the WI analysis synthesis system (with interpolation). The results of simulations and assessments, that indicate the performance of WI analysis synthesis system (standard) and WI analysis synthesis system (with interpolation) were presented with some speech signals. For assessment of the WI systems, the PESQ method were utilized to verify the perceptual quality of speech, and the SNRSEG-NCCF method, developed on this work, to verify the reconstruction of the waveforms and the phase delay per segment between the original and reconstructed speech signals. The WI analysis synthesis system (standard) presented good reconstruction of signal shown by the similarity between the time envelopes and the waveforms, and for the SNRSEG-NCCF measures. The WI system (standard) also presented a smaller phase displacement than the WI analysis synthesis system (with interpolation) and the perceptual quality of reconstructed speech ranged between "satisfactory" and "good" located on the band "communication quality" according the MOS rating. The preliminary version of the WI analysis synthesis system (with interpolation) presented better waveform reconstruction by segment and, for some parts of the signals, a better perceptual quality of speech than the WI analysis synthesis systems (standard), but with a larger phase displacement by segment between the originals and the reconstructed speech signals. In general the quality of reconstructed speech at the WI system (with interpolation) was lower than the WI system (standard), but the results indicate that the model has potential to present enhanced results by adjusting the interpolation process to improve the synchrony of signals.

ASSUNTO(S)

codificação da fala por interpolação de ondas engenharia eletrica speech processing waveform interpolation speech coding wi analysis sinthesis of speech codificação da fala análise e síntese da fala por interpolação de ondas codificador de voz waveform interpolation wi technique processamento da fala técnica wi processamento de sinais speech coding wi speech coding

ACESSO AO ARTIGO

Documentos Relacionados