Reformulação de um sistema de tratamento e reconhecimento de padrões para uma microbalança de quartzo utilizando redes neurais artificiais

Abstract

The classification of the coffee for the drink is made by "cup test", being difficult and delicate, therefore it suffers personal influence from the classifier. An alternative to decide this problem is the use of Artificial Neural Nets, as pattern recognition, to compose an artificial nose capable to evaluate the coffee odors. This work presents a system of treatment and recognition of coffee odors, using one technique of normalization for the pay-processing of the entrance data and nets Multi-Layer Perceptron(MLP), with different topologies, trained with backpropagation standard and backpropagation with momentum.