Enhancement of classification performance of an electronic nose using short-time Fourier transform

Abstract

This paper describes a method for enhancing classification performance of an electronic nose (E-nose) when measuring odors or flavors in ambient air. The method introduces short-time Fourier transform (STFT) to analyze the frequency characteristic of sensor response. The response of a gas sensor when exposed to an odor in ambient air, which is not in a closed system such as a chamber or sample headspace, is usually fluctuating due to odor concentration change caused by wind. The feature vectors of odor samples are created by using properly-selected frequency components. The results of principal component analysis (PCA) to the feature vectors indicate that the proposed feature extraction method can enhance the odor classification performance of electronic nose when used for measuring odors in ambient air.