A denoising method based on PCA and ICA in electronic nose for gases quantification

Due to the sensitivity of metal oxide gas sensors and complexity of sampling environments, the electronic nose (enose) measurement signal based on metal oxide sensor is sensitive to many disturbances. The noise interferences will influence the quantification accuracy of pattern recognition by an enose, and then reduce the practical monitoring accuracy. This paper proposed a hybrid denoising algorithm based on principle component analysis (PCA) reconstruction and independent component analysis (ICA). By extracting the true signal features which can fully reflect the concentration information from the original feature information, and realize the concentration estimation of indoor formaldehyde and benzene combined with radial basis function (RBF) neural network. Experimental results demonstrate that the proposed method in this paper can improve the prediction accuracy and robustness of an enose when it is applied for enose signal feature extraction and noise interference elimination.