Software Analysis Techniques For Odor Analysis and Classification Using the Electronic Nose

Abstract

The objectives of this thesis were to compare methods of feature extraction and data classification used in electronic nose. The NC State electronic nose (e-nose) was used to discriminate between SkipJack tuna (Katsuwonus pelamis) samples cooked at three temperatures: raw, heated to 55°C, and heated to 85°C. The thirty-six samples were analyzed by the e-nose on three separate days. The data were combined into one large set and randomly divided into a training (60%) and a testing (40%) set. The samples were labeled according to cooking treatment. Linear Discriminant Analysis (LDA) and Principal Component Analysis (PCA) were used for feature extraction. Extracted features from the training and testing sets were used to achieve a classification percentage using Least Squares (LS) and K-Nearest Neighbor (KNN). Data from a bell integral were used to train a feed-forward Artificial Neural Network (ANN) with a backpropagation algorithm. LDA proved to be a better method of feature extraction than PCA. ANN performance was not statistically different from LS, and performed better than KNN, with PCA as feature extraction. Both KNN and LS using LDA as feature extraction outperformed the ANN and the same methods using PCA.