Untitled in english

Author(s)

Miguel Antonio Fernandes Soler

• Document type: Master's Dissertation
• Press: São Paulo.
• Institution: Universidade de São Paulo (USP). Escola Politécnica (EP/BC)
• Defense date: 1998-12-22

Examining board members

Euvaldo Ferreira Cabral Junior; Jose Carlos Teixeira de Barros Moraes; Fabio Violaro

• Advisor: Euvaldo Ferreira Cabral Junior

Abstract

This work accomplished a study of chaotic systems and artificial neural networks aiming the application on voice criptophony systems. Some steps have been followed: selection of the chaotic systems, implementation of artificial neural networks for the class recognition, creation of patterns for classification, creation of a mixed signal for the transmitter, and the implementation of a criptophony system prototype. For the generation of the temporal series to be mixed with the voice signal to be sent, piece-wise ergodic linear chaotic maps have been chosen, which present characteristics that fit for the task, like, for instance, characteristics of pseudo-random systems. Tests have been performed with the Multi-Layer Perceptron, and the Probabilistic Neural Network. The MLP presented the best results. The signal with the voice information was mixed with other signals from chaotic maps, producing the signal to be transmitted. The voice criptophony system was efficiently implemented in both software and hardware. This work gives then giving a practical alternative to signal transmission in security systems, and allows a theoretical and practical basis for future developments and improvements. (AU)