Utilização de Técnicas de Aprendizado de Máquina para Demodulação de Sinais

The maturity of software Defined Radio (SDR) technologies associated with the greater availability of machine learning tools have favored the study of new proposals for communication systems. In this context, this article discusses the use of machine learning techniques in the construction of a digital demodulator Quadrature Phase Shift Keying (QPSK) using a set of training and validation data generated from GNU Radio. Unlike traditional methods, this demodulator does not depend on previous considerations about channel models, but uses its data set to carry out the supervised training of the demodulator symbol constellation for different environments, obtaining a performance comparable or superior to traditional demodulators.