Efficiency of gas detection algorithms using fluctuation enhanced sensing

Efficiency of various gas detection algorithms by applying fluctuation enhanced sensing method was discussed. We have analyzed resistance noise observed in resistive WO₃-nanowires gas sensing layers. Power spectral densities of the recorded noise were used as the input data vectors for two algorithms: the principal component analysis (PCA) and the support vector machine (SVM). The data were used to determine gas concentration by regression methods. Additionally, the SVM algorithm used the slope of 1/f noise estimated for consecutive low frequency bands only to reduce the volume of computation by limiting the input data vector size. The results show that the SVM method gives the best results when the input vector is a power spectral density.