Enhancing multiple classifier system performance for machine olfaction using odor-type signatures

Traditional odor classification systems used in machine olfaction devices, which are often called electronic noses, are implemented independently for each application dataset. For different types of odor samples (dissimilar odor datasets), some researchers have proposed a multiple classifier system that combines or fuses the classification outputs of individual, independent classifiers designed specifically for each dissimilar odor. However, in this approach, the classification system has to be reconstructed when new dissimilar odors are added to the machine's operation. Moreover, the multiple classifier system's performance is likely to be degraded due to the added complexity of the combined system. In this study, an approach to assign an unknown odor sample to one specific classifier in the multiple classifier set is proposed that is based on an odor-type signature derived from the sensor array's response waveforms. This novel approach enables an independent design of the classifier for each dissimilar odor, which is very useful when new odors need to be added to an existing machine olfaction system.