Reconhecimento do orador

Abstract

Verification of a person identity based in voice alone is an evolving technology with applications in secure access to information, access control to buildings, and telecommunications in general. This work addresses several of the possibilities available in the implementation of an automatic speaker verification system, independent of text, based in phoneme recognition. We studied which parameters to use, their stability over time, and also the phonemes to use for better performance. In this work we used feedforward neural networks in the verification process.