A classification of different odors by using artificial neural networks and fuzzy logic methods

In this thesis, it is tried to classify smells of mint, lemon, egg, rotten egg, angelica root, nail polish, naphthalene, rose water, and acetone, with electronic nose which is commonly used in industrial, military, environmental and medical applications. Artificial Neural Networks and Fuzzy Logic methods are used in pattern recognition part. Electronic nose which has been used in this work is designed in Department of Electrical and Electronics Engineering, Karadeniz Technical University. In this work, it is tried to increase accuracy rates of classification of data obtained from the electronic nose by using Fuzzy Logic methods which have not been used before, and by using Artificial Neural Networks which have been used before. In the study the change of classification performance is investigated by using various activation functions and architectures in Artificial Neural Network's hidden and output layers. In addition to Artificial Neural Networks, Fuzzy logic methods are used in this study, and effects of membership functions to classification accuracy is investigated. In order to determine classification performance, 5-fold cross validation method is used in pattern recognition part of electronic nose. Consequently the performance of these two methods is analysed and results are given.