Separação cega de fontes aplicada à sinais de áudio e imagem.

Abstract

The problem of overlapped signals has been studied since the very early days of telecommunications, either to minimize the e ects of noise, or simply to isolate a reference source. This work aims to study the source separation when there is no information of how those signals were mixed, known as blind source separation process. It will be analyzed the method for the separation of linear mixtures of independent sources through independent component analysis. Finally, it will be described the FastICA algorithm and evaluated its application to audio signals and images.