Online concentration independent feature dimension reduction of metal oxide gas sensor based E-Nose

In this paper, we describe an online dimension reduction of feature vector of metal oxide semiconductor (MOS) based electronic nose, operated in static temperature measurement for identification of gas. The sample space of gas consists of eight chemicals. We have used twelve number of MOS gas sensors, manufactured by Figaro sensors, Japan. The response of the sensors to the volatile organic compounds (VOC's) at different concentration of the VOC's were acquired. The pattern of response of MOS sensors were normalized to produce a consistent pattern independent of concentration. The native dimension reduction technique, principal component analysis (PCA) is being used for dimension reduction. The reduced feature vector is used as features for the pattern classifier based on neural network. Result clearly indicates that the VOC's can be distinguished using the reduced feature set.