Gas quantitative analysis by combining kernel independent component analysis and least squares support vector machine

Gas sensor array is an important part of electronic nose. The recognition ability of electronic nose is affected by the cross sensitivity of gas sensor array. We propose a pattern recognition method for electronic nose by combined use of kernel independent component analysis (KICA) and least squares support vector machine (LS-SVM) in this paper. In the proposed method, the KICA algorithm based on an entire function space of nonlinear subspace is firstly used for preprocessing gas sensor data in order to reduce the data correlation, and then a LS-SVM carries out the gas recognition. The measuring data was obtained from a gas of butane and ethanol for experiments. The results indicate that the proposed technique is effective in gas quantitative analysis, and gets higher precision than traditional techniques.