Signal processing inspired from the olfactory bulb for electronic noses

A bio-inspired signal processing method is proposed for electronic noses (e-noses). The proposed method contains an olfactory bulb model and a feature generation step. The structure of the olfactory bulb model is similar to the anatomical structure of mammals' olfactory bulb. It consists of olfactory receptor neurons, mitral cells, granule cells, periglomerular cells, and short axon cells. This model uses gas sensors' original response curves and transforms them to neuron spiking series no matter what kind the response curve is. This largely simplifies the follow-up feature generation step. Recurrence quantification analysis is employed to perform feature generation and the five most important features are selected. Finally, in order to verify the performance of the proposed method, seven kinds of Chinese liquors are tested and three classification methods are used to classify them. The experimental results demonstrate that the proposed method has a higher classification rate (99.05%) and also a steadier performance with the change of sensor number and types than the classic one.