Classification of odorants in the vapor phase using composite features for a portable e-nose system

We present an effective portable e-nose system that performs well even in noisy environments. Considering the characteristics of the e-nose data, we use an image covariance matrix-based method for extracting discriminant features for vapor classification. To construct composite vectors, primitive variables of the data measured by a sensor array are rearranged. Then, composite features are extracted by utilizing the information about the statistical dependency among multiple primitive variables, and a classifier for vapor classification is designed with these composite features. Experimental results with different volatile organic compounds data show that the proposed system has better classification performance than other methods in a noisy environment.