Journals Proceedings

Abstract

Present study deals the development of data fusion based artificial intelligence unit for the chemical sensor array based electronic nose (E-Nose) system. We focus particularly on feature level fusion of model surface acoustic wave (SAW) sensor array response for chemical class identification of volatile organic compounds (VOCs). Three methods are used for feature extraction namely: principal component analysis (PCA); independent component analysis (ICA) and kernel principal component analysis (KPCA). Fused features are generated with three unsupervised fusion schemes and validated in combination with support vector machine (SVM) classifier. Study is concluded by the analysis of 12 model SAW sensor array data sets. It suggests that amongst the three feature fusion schemes; feature fusion by summation result highest class recognition rate of VOCs.