Filtragem inversa para identificação do sinal glotal via sintese de fala

Abstract

The speech is a signal resulting from the modulation performed by the oral cavity and the supraglottal tract over the glottal signal. The glottal signal is an important signal because it brings information about the larynx, and can be used for the identification of laryngeal pathologies. To extract the glottal signal from the speech signal, it is necessary to use a system identification tool, the neural network. This paper aims to build two configurations from neural network architectures to extract the glottal signal from the speech signal and evaluate the performance of each one of them.