Study of a signal classification method in energy-saving electronic noses based on PCA and a periodic stochastic resonance

摘要

A new classification method based on principal component analysis (PCA) and aperiodic stochastic resonance (ASR) was presented in this paper. On the basis of analyzing the FHN model, the method of "maximum cross-correlation coefficient" was used to classify the energy-saving electronic nose's data. In the experiment, eight gas sensors were chosen to compose the energy-saving electronic nose's array, which was used to gather different level mildew data of read bean samples. These gathered data were reduced dimensions by PCA and then were passed through the ASR system to identify their categories. The read bean "maximum cross-correlation coefficients" of different mildew levels were found to be completely different, and the coefficient was almost a constant for each category, so it can be represented the different class accurately. The experimental results showed that the method of PCA and ASR in classifying energy-saving electronic nose's data was feasible, which provided a new way for the recognition of sensor array signals.