Arquitectura algorítmica para el reconocimiento de patrones fonéticos del habla sub-vocal en el español

Abstract

Silent speech is a recent subject in research. The objective is to allow the communication within high-noise environments, help people with language and speech disorders and under-water communication. Silent Speech is based on the information produced when a word is thought by a speaker-person but sonorous response is not produced. It means that the produced signals represent the intention of speak before the overt production (sonorous production). In this work is described the process of design and implementation of a recognition process for silent phonemes of the Spanish language. First was performed the acquisition of silent speech signals which were processed in further steps for feature extraction, using methodologies based in time and frequency representations. For the pattern recognition phase, algorithms based on Artificial Intelligence were implemented to perform the identification and classification of the silent speech signals. As result of this work were evaluated different processing and classification techniques with recognition indices near from 90%.