Título: Cancelación de Ruido a través de Técnicas Neurales
Autores: Montero, Fidel Ernesto Hernández; Urquiaga, Wilfredo Falcón
Resumo: The subject of this work is to apply neural techniques to the noise cancellation in a transmission line (e.g., return line of sensors, phone line). Basically, several models of Artificial Neural Networks (ANN) are developed, trying to solve the subject and the obtained results in each one of them are compared. The application of these nets is based, in principle, in that during the process of training, the neural architecture can learn the statistic of certain aleatory signal (e.g., noise). Then, later it will be possible to extract this noise from certain polluted useful signal when presenting this combination of signals (i.e., signal useful plus noise) to the input of the network and to act the network as a noise pattern “recognizer”, isolating this noise from the useful sign. During the investigation, two samples of noise were dealt: a sample of stationary noise (Gaussian white noise) and a sample of non-stationary noise (impulsive noise). The index employed to check the effectiveness of the operation of the ANN was the correlation between the useful signal without contamination and the signal obtained to the output of the neural network when the useful signal contaminated by the noise is in the input.
Artigo em pdf: 4cbrn_001.pdf
Arquivo BibTex: 4cbrn_001.bib